FFmpeg源代码简单分析-其他-libswscale的sws_scale()

参考链接

FFmpeg源代码简单分析：libswscale的sws_scale()_雷霄骅的博客-CSDN博客_bad dst image pointers

libswscale的sws_scale()

FFmpeg的图像处理（缩放，YUV/RGB格式转换）类库libswsscale中的sws_scale()函数。
libswscale是一个主要用于处理图片像素数据的类库。可以完成图片像素格式的转换，图片的拉伸等工作。
该类库常用的函数数量很少，一般情况下就3个：
- sws_getContext()：初始化一个SwsContext。
- sws_scale()：处理图像数据。
- sws_freeContext()：释放一个SwsContext。

Libswscale处理数据流程

Libswscale处理像素数据的流程可以概括为下图

从图中可以看出，libswscale处理数据有两条最主要的方式：unscaled和scaled。
unscaled用于处理不需要拉伸的像素数据（属于比较特殊的情况），scaled用于处理需要拉伸的像素数据。
Unscaled只需要对图像像素格式进行转换；而Scaled则除了对像素格式进行转换之外，还需要对图像进行缩放。
Scaled方式可以分成以下几个步骤：
- XXX to YUV Converter：首先将数据像素数据转换为8bitYUV格式；
- Horizontal scaler：水平拉伸图像，并且转换为15bitYUV；
- Vertical scaler：垂直拉伸图像；
- Output converter：转换为输出像素格式。

sws_scale()

sws_scale()是用于转换像素的函数。它的声明位于libswscale\swscale.h，如下所示。


/**
 * Scale the image slice in srcSlice and put the resulting scaled
 * slice in the image in dst. A slice is a sequence of consecutive
 * rows in an image.
 *
 * Slices have to be provided in sequential order, either in
 * top-bottom or bottom-top order. If slices are provided in
 * non-sequential order the behavior of the function is undefined.
 *
 * @param c         the scaling context previously created with
 *                  sws_getContext()
 * @param srcSlice  the array containing the pointers to the planes of
 *                  the source slice
 * @param srcStride the array containing the strides for each plane of
 *                  the source image
 * @param srcSliceY the position in the source image of the slice to
 *                  process, that is the number (counted starting from
 *                  zero) in the image of the first row of the slice
 * @param srcSliceH the height of the source slice, that is the number
 *                  of rows in the slice
 * @param dst       the array containing the pointers to the planes of
 *                  the destination image
 * @param dstStride the array containing the strides for each plane of
 *                  the destination image
 * @return          the height of the output slice
 */
int sws_scale(struct SwsContext *c, const uint8_t *const srcSlice[],
              const int srcStride[], int srcSliceY, int srcSliceH,
              uint8_t *const dst[], const int dstStride[]);

sws_scale()的定义位于libswscale\swscale.c，如下所示。


/**
 * swscale wrapper, so we don't need to export the SwsContext.
 * Assumes planar YUV to be in YUV order instead of YVU.
 */
int attribute_align_arg sws_scale(struct SwsContext *c,
                                  const uint8_t * const srcSlice[],
                                  const int srcStride[], int srcSliceY,
                                  int srcSliceH, uint8_t *const dst[],
                                  const int dstStride[])
{
    if (c->nb_slice_ctx)
        c = c->slice_ctx[0];
 
    return scale_internal(c, srcSlice, srcStride, srcSliceY, srcSliceH,
                          dst, dstStride, 0, c->dstH);
}

sws_scale内部调用了scale_internal，scale_internal函数封装了sws_scale的大多数代码


static int scale_internal(SwsContext *c,
                          const uint8_t * const srcSlice[], const int srcStride[],
                          int srcSliceY, int srcSliceH,
                          uint8_t *const dstSlice[], const int dstStride[],
                          int dstSliceY, int dstSliceH)
{
    const int scale_dst = dstSliceY > 0 || dstSliceH < c->dstH;
    const int frame_start = scale_dst || !c->sliceDir;
    int i, ret;
    const uint8_t *src2[4];
    uint8_t *dst2[4];
    int macro_height_src = isBayer(c->srcFormat) ? 2 : (1 << c->chrSrcVSubSample);
    int macro_height_dst = isBayer(c->dstFormat) ? 2 : (1 << c->chrDstVSubSample);
    // copy strides, so they can safely be modified
    int srcStride2[4];
    int dstStride2[4];
    int srcSliceY_internal = srcSliceY;
 
    if (!srcStride || !dstStride || !dstSlice || !srcSlice) {
        av_log(c, AV_LOG_ERROR, "One of the input parameters to sws_scale() is NULL, please check the calling code\n");
        return AVERROR(EINVAL);
    }
 
    if ((srcSliceY  & (macro_height_src - 1)) ||
        ((srcSliceH & (macro_height_src - 1)) && srcSliceY + srcSliceH != c->srcH) ||
        srcSliceY + srcSliceH > c->srcH) {
        av_log(c, AV_LOG_ERROR, "Slice parameters %d, %d are invalid\n", srcSliceY, srcSliceH);
        return AVERROR(EINVAL);
    }
 
    if ((dstSliceY  & (macro_height_dst - 1)) ||
        ((dstSliceH & (macro_height_dst - 1)) && dstSliceY + dstSliceH != c->dstH) ||
        dstSliceY + dstSliceH > c->dstH) {
        av_log(c, AV_LOG_ERROR, "Slice parameters %d, %d are invalid\n", dstSliceY, dstSliceH);
        return AVERROR(EINVAL);
    }
 
    if (!check_image_pointers(srcSlice, c->srcFormat, srcStride)) {
        av_log(c, AV_LOG_ERROR, "bad src image pointers\n");
        return AVERROR(EINVAL);
    }
    if (!check_image_pointers((const uint8_t* const*)dstSlice, c->dstFormat, dstStride)) {
        av_log(c, AV_LOG_ERROR, "bad dst image pointers\n");
        return AVERROR(EINVAL);
    }
 
    // do not mess up sliceDir if we have a "trailing" 0-size slice
    if (srcSliceH == 0)
        return 0;
 
    if (c->gamma_flag && c->cascaded_context[0])
        return scale_gamma(c, srcSlice, srcStride, srcSliceY, srcSliceH,
                           dstSlice, dstStride, dstSliceY, dstSliceH);
 
    if (c->cascaded_context[0] && srcSliceY == 0 && srcSliceH == c->cascaded_context[0]->srcH)
        return scale_cascaded(c, srcSlice, srcStride, srcSliceY, srcSliceH,
                              dstSlice, dstStride, dstSliceY, dstSliceH);
 
    if (!srcSliceY && (c->flags & SWS_BITEXACT) && c->dither == SWS_DITHER_ED && c->dither_error[0])
        for (i = 0; i < 4; i++)
            memset(c->dither_error[i], 0, sizeof(c->dither_error[0][0]) * (c->dstW+2));
 
    if (usePal(c->srcFormat))
        update_palette(c, (const uint32_t *)srcSlice[1]);
 
    memcpy(src2,       srcSlice,  sizeof(src2));
    memcpy(dst2,       dstSlice,  sizeof(dst2));
    memcpy(srcStride2, srcStride, sizeof(srcStride2));
    memcpy(dstStride2, dstStride, sizeof(dstStride2));
 
    if (frame_start && !scale_dst) {
        if (srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {
            av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");
            return AVERROR(EINVAL);
        }
 
        c->sliceDir = (srcSliceY == 0) ? 1 : -1;
    } else if (scale_dst)
        c->sliceDir = 1;
 
    if (c->src0Alpha && !c->dst0Alpha && isALPHA(c->dstFormat)) {
        uint8_t *base;
        int x,y;
 
        av_fast_malloc(&c->rgb0_scratch, &c->rgb0_scratch_allocated,
                       FFABS(srcStride[0]) * srcSliceH + 32);
        if (!c->rgb0_scratch)
            return AVERROR(ENOMEM);
 
        base = srcStride[0] < 0 ? c->rgb0_scratch - srcStride[0] * (srcSliceH-1) :
                                  c->rgb0_scratch;
        for (y=0; y<srcSliceH; y++){
            memcpy(base + srcStride[0]*y, src2[0] + srcStride[0]*y, 4*c->srcW);
            for (x=c->src0Alpha-1; x<4*c->srcW; x+=4) {
                base[ srcStride[0]*y + x] = 0xFF;
            }
        }
        src2[0] = base;
    }
 
    if (c->srcXYZ && !(c->dstXYZ && c->srcW==c->dstW && c->srcH==c->dstH)) {
        uint8_t *base;
 
        av_fast_malloc(&c->xyz_scratch, &c->xyz_scratch_allocated,
                       FFABS(srcStride[0]) * srcSliceH + 32);
        if (!c->xyz_scratch)
            return AVERROR(ENOMEM);
 
        base = srcStride[0] < 0 ? c->xyz_scratch - srcStride[0] * (srcSliceH-1) :
                                  c->xyz_scratch;
 
        xyz12Torgb48(c, (uint16_t*)base, (const uint16_t*)src2[0], srcStride[0]/2, srcSliceH);
        src2[0] = base;
    }
 
    if (c->sliceDir != 1) {
        // slices go from bottom to top => we flip the image internally
        for (i=0; i<4; i++) {
            srcStride2[i] *= -1;
            dstStride2[i] *= -1;
        }
 
        src2[0] += (srcSliceH - 1) * srcStride[0];
        if (!usePal(c->srcFormat))
            src2[1] += ((srcSliceH >> c->chrSrcVSubSample) - 1) * srcStride[1];
        src2[2] += ((srcSliceH >> c->chrSrcVSubSample) - 1) * srcStride[2];
        src2[3] += (srcSliceH - 1) * srcStride[3];
        dst2[0] += ( c->dstH                         - 1) * dstStride[0];
        dst2[1] += ((c->dstH >> c->chrDstVSubSample) - 1) * dstStride[1];
        dst2[2] += ((c->dstH >> c->chrDstVSubSample) - 1) * dstStride[2];
        dst2[3] += ( c->dstH                         - 1) * dstStride[3];
 
        srcSliceY_internal = c->srcH-srcSliceY-srcSliceH;
    }
    reset_ptr(src2, c->srcFormat);
    reset_ptr((void*)dst2, c->dstFormat);
 
    if (c->convert_unscaled) {
        int offset  = srcSliceY_internal;
        int slice_h = srcSliceH;
 
        // for dst slice scaling, offset the pointers to match the unscaled API
        if (scale_dst) {
            av_assert0(offset == 0);
            for (i = 0; i < 4 && src2[i]; i++) {
                if (!src2[i] || (i > 0 && usePal(c->srcFormat)))
                    break;
                src2[i] += (dstSliceY >> ((i == 1 || i == 2) ? c->chrSrcVSubSample : 0)) * srcStride2[i];
            }
 
            for (i = 0; i < 4 && dst2[i]; i++) {
                if (!dst2[i] || (i > 0 && usePal(c->dstFormat)))
                    break;
                dst2[i] -= (dstSliceY >> ((i == 1 || i == 2) ? c->chrDstVSubSample : 0)) * dstStride2[i];
            }
            offset  = dstSliceY;
            slice_h = dstSliceH;
        }
 
        ret = c->convert_unscaled(c, src2, srcStride2, offset, slice_h,
                                  dst2, dstStride2);
        if (scale_dst)
            dst2[0] += dstSliceY * dstStride2[0];
    } else {
        ret = swscale(c, src2, srcStride2, srcSliceY_internal, srcSliceH,
                      dst2, dstStride2, dstSliceY, dstSliceH);
    }
 
    if (c->dstXYZ && !(c->srcXYZ && c->srcW==c->dstW && c->srcH==c->dstH)) {
        uint16_t *dst16;
 
        if (scale_dst) {
            dst16 = (uint16_t *)dst2[0];
        } else {
            int dstY = c->dstY ? c->dstY : srcSliceY + srcSliceH;
 
            av_assert0(dstY >= ret);
            av_assert0(ret >= 0);
            av_assert0(c->dstH >= dstY);
            dst16 = (uint16_t*)(dst2[0] + (dstY - ret) * dstStride2[0]);
        }
 
        /* replace on the same data */
        rgb48Toxyz12(c, dst16, dst16, dstStride2[0]/2, ret);
    }
 
    /* reset slice direction at end of frame */
    if ((srcSliceY_internal + srcSliceH == c->srcH) || scale_dst)
        c->sliceDir = 0;
 
    return ret;
}

从sws_scale()的定义可以看出，它封装了SwsContext中的swscale()（注意这个函数中间没有“_”）。函数最重要的一句代码就是“swscale()”。
除此之外，函数还做了一些增加“兼容性”的一些处理。
函数的主要步骤如下所示。

1.检查输入的图像参数的合理性。

这一步骤首先检查输入输出的参数是否为空，然后通过调用check_image_pointers()检查输入输出图像的内存是否正确分配。
check_image_pointers()的定义如下所示。


static int check_image_pointers(const uint8_t * const data[4], enum AVPixelFormat pix_fmt,
                                const int linesizes[4])
{
    const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
    int i;
 
    av_assert2(desc);
 
    for (i = 0; i < 4; i++) {
        int plane = desc->comp[i].plane;
        if (!data[plane] || !linesizes[plane])
            return 0;
    }
 
    return 1;
}

从check_image_pointers()的定义可以看出，在特定像素格式前提下，如果该像素格式应该包含像素的分量为空，就返回0，否则返回1。

2.如果输入像素数据中使用了“调色板”（palette），则进行一些相应的处理。

这一步通过函数usePal()来判定。
usePal()的定义如下。


static av_always_inline int usePal(enum AVPixelFormat pix_fmt)
{
    switch (pix_fmt) {
    case AV_PIX_FMT_PAL8:
    case AV_PIX_FMT_BGR4_BYTE:
    case AV_PIX_FMT_BGR8:
    case AV_PIX_FMT_GRAY8:
    case AV_PIX_FMT_RGB4_BYTE:
    case AV_PIX_FMT_RGB8:
        return 1;
    default:
        return 0;
    }
}

3.其它一些特殊格式的处理，比如说Alpha，XYZ等的处理（这方面没有研究过）。
4.如果输入的图像的扫描方式是从底部到顶部的（一般情况下是从顶部到底部），则将图像进行反转。
5.调用SwsContext中的swscale()。
SwsContext中的swscale()

swscale这个变量的类型是SwsFunc，实际上就是一个函数指针。它是整个类库的核心。当我们从外部调用swscale()函数的时候，实际上就是调用了SwsContext中的这个名称为swscale的变量（注意外部函数接口和这个内部函数指针的名字是一样的，但不是一回事）。
可以看一下SwsFunc这个类型的定义：


typedef int (*SwsFunc)(struct SwsContext *context, const uint8_t *src[],
                       int srcStride[], int srcSliceY, int srcSliceH,
                       uint8_t *dst[], int dstStride[]);

可以看出SwsFunc的定义的参数类型和libswscale类库外部接口函数swscale()的参数类型一模一样。
在libswscale中，该指针的指向可以分成2种情况：
- 1.图像没有伸缩的时候，指向专有的像素转换函数
- 2.图像有伸缩的时候，指向swscale()函数。
在调用sws_getContext()初始化SwsContext的时候，会在其子函数sws_init_context()中对swscale指针进行赋值。如果图像没有进行拉伸，则会调用ff_get_unscaled_swscale()对其进行赋值；如果图像进行了拉伸，则会调用ff_getSwsFunc()对其进行赋值。
下面分别看一下这2种情况。

没有拉伸--专有的像素转换函数

如果图像没有进行拉伸，则会调用ff_get_unscaled_swscale()对SwsContext的swscale进行赋值。
上篇文章中记录了这个函数，在这里回顾一下。

ff_get_unscaled_swscale()

ff_get_unscaled_swscale()的定义如下。


void ff_get_unscaled_swscale(SwsContext *c)
{
    const enum AVPixelFormat srcFormat = c->srcFormat;
    const enum AVPixelFormat dstFormat = c->dstFormat;
    const int flags = c->flags;
    const int dstH = c->dstH;
    const int dstW = c->dstW;
    int needsDither;
 
    needsDither = isAnyRGB(dstFormat) &&
            c->dstFormatBpp < 24 &&
           (c->dstFormatBpp < c->srcFormatBpp || (!isAnyRGB(srcFormat)));
 
    /* yv12_to_nv12 */
    if ((srcFormat == AV_PIX_FMT_YUV420P || srcFormat == AV_PIX_FMT_YUVA420P) &&
        (dstFormat == AV_PIX_FMT_NV12 || dstFormat == AV_PIX_FMT_NV21)) {
        c->convert_unscaled = planarToNv12Wrapper;
    }
    /* yv24_to_nv24 */
    if ((srcFormat == AV_PIX_FMT_YUV444P || srcFormat == AV_PIX_FMT_YUVA444P) &&
        (dstFormat == AV_PIX_FMT_NV24 || dstFormat == AV_PIX_FMT_NV42)) {
        c->convert_unscaled = planarToNv24Wrapper;
    }
    /* nv12_to_yv12 */
    if (dstFormat == AV_PIX_FMT_YUV420P &&
        (srcFormat == AV_PIX_FMT_NV12 || srcFormat == AV_PIX_FMT_NV21)) {
        c->convert_unscaled = nv12ToPlanarWrapper;
    }
    /* nv24_to_yv24 */
    if (dstFormat == AV_PIX_FMT_YUV444P &&
        (srcFormat == AV_PIX_FMT_NV24 || srcFormat == AV_PIX_FMT_NV42)) {
        c->convert_unscaled = nv24ToPlanarWrapper;
    }
    /* yuv2bgr */
    if ((srcFormat == AV_PIX_FMT_YUV420P || srcFormat == AV_PIX_FMT_YUV422P ||
         srcFormat == AV_PIX_FMT_YUVA420P) && isAnyRGB(dstFormat) &&
        !(flags & SWS_ACCURATE_RND) && (c->dither == SWS_DITHER_BAYER || c->dither == SWS_DITHER_AUTO) && !(dstH & 1)) {
        c->convert_unscaled = ff_yuv2rgb_get_func_ptr(c);
        c->dst_slice_align = 2;
    }
    /* yuv420p1x_to_p01x */
    if ((srcFormat == AV_PIX_FMT_YUV420P10 || srcFormat == AV_PIX_FMT_YUVA420P10 ||
         srcFormat == AV_PIX_FMT_YUV420P12 ||
         srcFormat == AV_PIX_FMT_YUV420P14 ||
         srcFormat == AV_PIX_FMT_YUV420P16 || srcFormat == AV_PIX_FMT_YUVA420P16) &&
        (dstFormat == AV_PIX_FMT_P010 || dstFormat == AV_PIX_FMT_P016)) {
        c->convert_unscaled = planarToP01xWrapper;
    }
    /* yuv420p_to_p01xle */
    if ((srcFormat == AV_PIX_FMT_YUV420P || srcFormat == AV_PIX_FMT_YUVA420P) &&
        (dstFormat == AV_PIX_FMT_P010LE || dstFormat == AV_PIX_FMT_P016LE)) {
        c->convert_unscaled = planar8ToP01xleWrapper;
    }
 
    if (srcFormat == AV_PIX_FMT_YUV410P && !(dstH & 3) &&
        (dstFormat == AV_PIX_FMT_YUV420P || dstFormat == AV_PIX_FMT_YUVA420P) &&
        !(flags & SWS_BITEXACT)) {
        c->convert_unscaled = yvu9ToYv12Wrapper;
        c->dst_slice_align = 4;
    }
 
    /* bgr24toYV12 */
    if (srcFormat == AV_PIX_FMT_BGR24 &&
        (dstFormat == AV_PIX_FMT_YUV420P || dstFormat == AV_PIX_FMT_YUVA420P) &&
        !(flags & SWS_ACCURATE_RND) && !(dstW&1))
        c->convert_unscaled = bgr24ToYv12Wrapper;
 
    /* RGB/BGR -> RGB/BGR (no dither needed forms) */
    if (isAnyRGB(srcFormat) && isAnyRGB(dstFormat) && findRgbConvFn(c)
        && (!needsDither || (c->flags&(SWS_FAST_BILINEAR|SWS_POINT))))
        c->convert_unscaled = rgbToRgbWrapper;
 
    /* RGB to planar RGB */
    if ((srcFormat == AV_PIX_FMT_GBRP && dstFormat == AV_PIX_FMT_GBRAP) ||
        (srcFormat == AV_PIX_FMT_GBRAP && dstFormat == AV_PIX_FMT_GBRP))
        c->convert_unscaled = planarRgbToplanarRgbWrapper;
 
#define isByteRGB(f) (             \
        f == AV_PIX_FMT_RGB32   || \
        f == AV_PIX_FMT_RGB32_1 || \
        f == AV_PIX_FMT_RGB24   || \
        f == AV_PIX_FMT_BGR32   || \
        f == AV_PIX_FMT_BGR32_1 || \
        f == AV_PIX_FMT_BGR24)
 
    if (srcFormat == AV_PIX_FMT_GBRP && isPlanar(srcFormat) && isByteRGB(dstFormat))
        c->convert_unscaled = planarRgbToRgbWrapper;
 
    if (srcFormat == AV_PIX_FMT_GBRAP && isByteRGB(dstFormat))
        c->convert_unscaled = planarRgbaToRgbWrapper;
 
    if ((srcFormat == AV_PIX_FMT_RGB48LE  || srcFormat == AV_PIX_FMT_RGB48BE  ||
         srcFormat == AV_PIX_FMT_BGR48LE  || srcFormat == AV_PIX_FMT_BGR48BE  ||
         srcFormat == AV_PIX_FMT_RGBA64LE || srcFormat == AV_PIX_FMT_RGBA64BE ||
         srcFormat == AV_PIX_FMT_BGRA64LE || srcFormat == AV_PIX_FMT_BGRA64BE) &&
        (dstFormat == AV_PIX_FMT_GBRP9LE  || dstFormat == AV_PIX_FMT_GBRP9BE  ||
         dstFormat == AV_PIX_FMT_GBRP10LE || dstFormat == AV_PIX_FMT_GBRP10BE ||
         dstFormat == AV_PIX_FMT_GBRP12LE || dstFormat == AV_PIX_FMT_GBRP12BE ||
         dstFormat == AV_PIX_FMT_GBRP14LE || dstFormat == AV_PIX_FMT_GBRP14BE ||
         dstFormat == AV_PIX_FMT_GBRP16LE || dstFormat == AV_PIX_FMT_GBRP16BE ||
         dstFormat == AV_PIX_FMT_GBRAP10LE || dstFormat == AV_PIX_FMT_GBRAP10BE ||
         dstFormat == AV_PIX_FMT_GBRAP12LE || dstFormat == AV_PIX_FMT_GBRAP12BE ||
         dstFormat == AV_PIX_FMT_GBRAP16LE || dstFormat == AV_PIX_FMT_GBRAP16BE ))
        c->convert_unscaled = Rgb16ToPlanarRgb16Wrapper;
 
    if ((srcFormat == AV_PIX_FMT_GBRP9LE  || srcFormat == AV_PIX_FMT_GBRP9BE  ||
         srcFormat == AV_PIX_FMT_GBRP16LE || srcFormat == AV_PIX_FMT_GBRP16BE ||
         srcFormat == AV_PIX_FMT_GBRP10LE || srcFormat == AV_PIX_FMT_GBRP10BE ||
         srcFormat == AV_PIX_FMT_GBRP12LE || srcFormat == AV_PIX_FMT_GBRP12BE ||
         srcFormat == AV_PIX_FMT_GBRP14LE || srcFormat == AV_PIX_FMT_GBRP14BE ||
         srcFormat == AV_PIX_FMT_GBRAP10LE || srcFormat == AV_PIX_FMT_GBRAP10BE ||
         srcFormat == AV_PIX_FMT_GBRAP12LE || srcFormat == AV_PIX_FMT_GBRAP12BE ||
         srcFormat == AV_PIX_FMT_GBRAP16LE || srcFormat == AV_PIX_FMT_GBRAP16BE) &&
        (dstFormat == AV_PIX_FMT_RGB48LE  || dstFormat == AV_PIX_FMT_RGB48BE  ||
         dstFormat == AV_PIX_FMT_BGR48LE  || dstFormat == AV_PIX_FMT_BGR48BE  ||
         dstFormat == AV_PIX_FMT_RGBA64LE || dstFormat == AV_PIX_FMT_RGBA64BE ||
         dstFormat == AV_PIX_FMT_BGRA64LE || dstFormat == AV_PIX_FMT_BGRA64BE))
        c->convert_unscaled = planarRgb16ToRgb16Wrapper;
 
    if (av_pix_fmt_desc_get(srcFormat)->comp[0].depth == 8 &&
        isPackedRGB(srcFormat) && dstFormat == AV_PIX_FMT_GBRP)
        c->convert_unscaled = rgbToPlanarRgbWrapper;
 
    if (isBayer(srcFormat)) {
        if (dstFormat == AV_PIX_FMT_RGB24)
            c->convert_unscaled = bayer_to_rgb24_wrapper;
        else if (dstFormat == AV_PIX_FMT_RGB48)
            c->convert_unscaled = bayer_to_rgb48_wrapper;
        else if (dstFormat == AV_PIX_FMT_YUV420P)
            c->convert_unscaled = bayer_to_yv12_wrapper;
        else if (!isBayer(dstFormat)) {
            av_log(c, AV_LOG_ERROR, "unsupported bayer conversion\n");
            av_assert0(0);
        }
    }
 
    /* bswap 16 bits per pixel/component packed formats */
    if (IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_BAYER_BGGR16) ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_BAYER_RGGB16) ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_BAYER_GBRG16) ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_BAYER_GRBG16) ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_BGR444) ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_BGR48)  ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_BGR555) ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_BGR565) ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_BGRA64) ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_GRAY9)  ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_GRAY10) ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_GRAY12) ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_GRAY14) ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_GRAY16) ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_YA16)   ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_AYUV64) ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_GBRP9)  ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_GBRP10) ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_GBRP12) ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_GBRP14) ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_GBRP16) ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_GBRAP10) ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_GBRAP12) ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_GBRAP16) ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_RGB444) ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_RGB48)  ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_RGB555) ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_RGB565) ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_RGBA64) ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_XYZ12)  ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_YUV420P9)  ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_YUV420P10) ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_YUV420P12) ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_YUV420P14) ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_YUV420P16) ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_YUV422P9)  ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_YUV422P10) ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_YUV422P12) ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_YUV422P14) ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_YUV422P16) ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_YUV440P10) ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_YUV440P12) ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_YUV444P9)  ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_YUV444P10) ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_YUV444P12) ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_YUV444P14) ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_YUV444P16))
        c->convert_unscaled = bswap_16bpc;
 
    /* bswap 32 bits per pixel/component formats */
    if (IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_GBRPF32) ||
        IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_GBRAPF32))
        c->convert_unscaled = bswap_32bpc;
 
    if (usePal(srcFormat) && isByteRGB(dstFormat))
        c->convert_unscaled = palToRgbWrapper;
 
    if (srcFormat == AV_PIX_FMT_YUV422P) {
        if (dstFormat == AV_PIX_FMT_YUYV422)
            c->convert_unscaled = yuv422pToYuy2Wrapper;
        else if (dstFormat == AV_PIX_FMT_UYVY422)
            c->convert_unscaled = yuv422pToUyvyWrapper;
    }
 
    /* uint Y to float Y */
    if (srcFormat == AV_PIX_FMT_GRAY8 && dstFormat == AV_PIX_FMT_GRAYF32){
        c->convert_unscaled = uint_y_to_float_y_wrapper;
    }
 
    /* float Y to uint Y */
    if (srcFormat == AV_PIX_FMT_GRAYF32 && dstFormat == AV_PIX_FMT_GRAY8){
        c->convert_unscaled = float_y_to_uint_y_wrapper;
    }
 
    /* LQ converters if -sws 0 or -sws 4*/
    if (c->flags&(SWS_FAST_BILINEAR|SWS_POINT)) {
        /* yv12_to_yuy2 */
        if (srcFormat == AV_PIX_FMT_YUV420P || srcFormat == AV_PIX_FMT_YUVA420P) {
            if (dstFormat == AV_PIX_FMT_YUYV422)
                c->convert_unscaled = planarToYuy2Wrapper;
            else if (dstFormat == AV_PIX_FMT_UYVY422)
                c->convert_unscaled = planarToUyvyWrapper;
        }
    }
    if (srcFormat == AV_PIX_FMT_YUYV422 &&
       (dstFormat == AV_PIX_FMT_YUV420P || dstFormat == AV_PIX_FMT_YUVA420P))
        c->convert_unscaled = yuyvToYuv420Wrapper;
    if (srcFormat == AV_PIX_FMT_UYVY422 &&
       (dstFormat == AV_PIX_FMT_YUV420P || dstFormat == AV_PIX_FMT_YUVA420P))
        c->convert_unscaled = uyvyToYuv420Wrapper;
    if (srcFormat == AV_PIX_FMT_YUYV422 && dstFormat == AV_PIX_FMT_YUV422P)
        c->convert_unscaled = yuyvToYuv422Wrapper;
    if (srcFormat == AV_PIX_FMT_UYVY422 && dstFormat == AV_PIX_FMT_YUV422P)
        c->convert_unscaled = uyvyToYuv422Wrapper;
 
#define isPlanarGray(x) (isGray(x) && (x) != AV_PIX_FMT_YA8 && (x) != AV_PIX_FMT_YA16LE && (x) != AV_PIX_FMT_YA16BE)
    /* simple copy */
    if ( srcFormat == dstFormat ||
        (srcFormat == AV_PIX_FMT_YUVA420P && dstFormat == AV_PIX_FMT_YUV420P) ||
        (srcFormat == AV_PIX_FMT_YUV420P && dstFormat == AV_PIX_FMT_YUVA420P) ||
        (isFloat(srcFormat) == isFloat(dstFormat)) && ((isPlanarYUV(srcFormat) && isPlanarGray(dstFormat)) ||
        (isPlanarYUV(dstFormat) && isPlanarGray(srcFormat)) ||
        (isPlanarGray(dstFormat) && isPlanarGray(srcFormat)) ||
        (isPlanarYUV(srcFormat) && isPlanarYUV(dstFormat) &&
         c->chrDstHSubSample == c->chrSrcHSubSample &&
         c->chrDstVSubSample == c->chrSrcVSubSample &&
         !isSemiPlanarYUV(srcFormat) && !isSemiPlanarYUV(dstFormat))))
    {
        if (isPacked(c->srcFormat))
            c->convert_unscaled = packedCopyWrapper;
        else /* Planar YUV or gray */
            c->convert_unscaled = planarCopyWrapper;
    }
 
    if (ARCH_PPC)
        ff_get_unscaled_swscale_ppc(c);
     if (ARCH_ARM)
         ff_get_unscaled_swscale_arm(c);
    if (ARCH_AARCH64)
        ff_get_unscaled_swscale_aarch64(c);
}

从代码中可以看出，它根据输入输出像素格式的不同，选择了不同的转换函数。
例如YUV420P转换NV12的时候，就会将planarToNv12Wrapper()赋值给SwsContext的swscale指针。

有拉伸--swscale()

如果图像进行了拉伸，则会调用ff_getSwsFunc()对SwsContext的swscale进行赋值。
ff_getSwsFunc 函数被弃用
参考链接：FFmpeg源代码简单分析-其他-libswscale的sws_getContext()_MY CUP OF TEA的博客-CSDN博客
ff_sws_init_scale函数的内部执行逻辑和 ff_getSwsFunc 类似


void ff_sws_init_scale(SwsContext *c)
{
    sws_init_swscale(c);
 
    if (ARCH_PPC)
        ff_sws_init_swscale_ppc(c);
    if (ARCH_X86)
        ff_sws_init_swscale_x86(c);
    if (ARCH_AARCH64)
        ff_sws_init_swscale_aarch64(c);
    if (ARCH_ARM)
        ff_sws_init_swscale_arm(c);
}


static av_cold void sws_init_swscale(SwsContext *c)
{
    enum AVPixelFormat srcFormat = c->srcFormat;
 
    ff_sws_init_output_funcs(c, &c->yuv2plane1, &c->yuv2planeX,
                             &c->yuv2nv12cX, &c->yuv2packed1,
                             &c->yuv2packed2, &c->yuv2packedX, &c->yuv2anyX);
 
    ff_sws_init_input_funcs(c);
 
    if (c->srcBpc == 8) {
        if (c->dstBpc <= 14) {
            c->hyScale = c->hcScale = hScale8To15_c;
            if (c->flags & SWS_FAST_BILINEAR) {
                c->hyscale_fast = ff_hyscale_fast_c;
                c->hcscale_fast = ff_hcscale_fast_c;
            }
        } else {
            c->hyScale = c->hcScale = hScale8To19_c;
        }
    } else {
        c->hyScale = c->hcScale = c->dstBpc > 14 ? hScale16To19_c
                                                 : hScale16To15_c;
    }
 
    ff_sws_init_range_convert(c);
 
    if (!(isGray(srcFormat) || isGray(c->dstFormat) ||
          srcFormat == AV_PIX_FMT_MONOBLACK || srcFormat == AV_PIX_FMT_MONOWHITE))
        c->needs_hcscale = 1;
}

未找到代码依据

注意，sws_init_context()对SwsContext的swscale进行赋值的语句是：
c->swscale = ff_getSwsFunc(c);
即把ff_getSwsFunc()的返回值赋值给SwsContext的swscale指针；而ff_getSwsFunc()的返回值是一个静态函数，名称就叫做“swscale”。

下面我们看一下这个swscale()静态函数的定义。


static int swscale(SwsContext *c, const uint8_t *src[],
                   int srcStride[], int srcSliceY, int srcSliceH,
                   uint8_t *dst[], int dstStride[],
                   int dstSliceY, int dstSliceH)
{
    const int scale_dst = dstSliceY > 0 || dstSliceH < c->dstH;
 
    /* load a few things into local vars to make the code more readable?
     * and faster */
    const int dstW                   = c->dstW;
    int dstH                         = c->dstH;
 
    const enum AVPixelFormat dstFormat = c->dstFormat;
    const int flags                  = c->flags;
    int32_t *vLumFilterPos           = c->vLumFilterPos;
    int32_t *vChrFilterPos           = c->vChrFilterPos;
 
    const int vLumFilterSize         = c->vLumFilterSize;
    const int vChrFilterSize         = c->vChrFilterSize;
 
    yuv2planar1_fn yuv2plane1        = c->yuv2plane1;
    yuv2planarX_fn yuv2planeX        = c->yuv2planeX;
    yuv2interleavedX_fn yuv2nv12cX   = c->yuv2nv12cX;
    yuv2packed1_fn yuv2packed1       = c->yuv2packed1;
    yuv2packed2_fn yuv2packed2       = c->yuv2packed2;
    yuv2packedX_fn yuv2packedX       = c->yuv2packedX;
    yuv2anyX_fn yuv2anyX             = c->yuv2anyX;
    const int chrSrcSliceY           =                srcSliceY >> c->chrSrcVSubSample;
    const int chrSrcSliceH           = AV_CEIL_RSHIFT(srcSliceH,   c->chrSrcVSubSample);
    int should_dither                = isNBPS(c->srcFormat) ||
                                       is16BPS(c->srcFormat);
    int lastDstY;
 
    /* vars which will change and which we need to store back in the context */
    int dstY         = c->dstY;
    int lastInLumBuf = c->lastInLumBuf;
    int lastInChrBuf = c->lastInChrBuf;
 
    int lumStart = 0;
    int lumEnd = c->descIndex[0];
    int chrStart = lumEnd;
    int chrEnd = c->descIndex[1];
    int vStart = chrEnd;
    int vEnd = c->numDesc;
    SwsSlice *src_slice = &c->slice[lumStart];
    SwsSlice *hout_slice = &c->slice[c->numSlice-2];
    SwsSlice *vout_slice = &c->slice[c->numSlice-1];
    SwsFilterDescriptor *desc = c->desc;
 
    int needAlpha = c->needAlpha;
 
    int hasLumHoles = 1;
    int hasChrHoles = 1;
 
    if (isPacked(c->srcFormat)) {
        src[1] =
        src[2] =
        src[3] = src[0];
        srcStride[1] =
        srcStride[2] =
        srcStride[3] = srcStride[0];
    }
    srcStride[1] *= 1 << c->vChrDrop;
    srcStride[2] *= 1 << c->vChrDrop;
 
    DEBUG_BUFFERS("swscale() %p[%d] %p[%d] %p[%d] %p[%d] -> %p[%d] %p[%d] %p[%d] %p[%d]\n",
                  src[0], srcStride[0], src[1], srcStride[1],
                  src[2], srcStride[2], src[3], srcStride[3],
                  dst[0], dstStride[0], dst[1], dstStride[1],
                  dst[2], dstStride[2], dst[3], dstStride[3]);
    DEBUG_BUFFERS("srcSliceY: %d srcSliceH: %d dstY: %d dstH: %d\n",
                  srcSliceY, srcSliceH, dstY, dstH);
    DEBUG_BUFFERS("vLumFilterSize: %d vChrFilterSize: %d\n",
                  vLumFilterSize, vChrFilterSize);
 
    if (dstStride[0]&15 || dstStride[1]&15 ||
        dstStride[2]&15 || dstStride[3]&15) {
        SwsContext *const ctx = c->parent ? c->parent : c;
        if (flags & SWS_PRINT_INFO &&
            !atomic_exchange_explicit(&ctx->stride_unaligned_warned, 1, memory_order_relaxed)) {
            av_log(c, AV_LOG_WARNING,
                   "Warning: dstStride is not aligned!\n"
                   "         ->cannot do aligned memory accesses anymore\n");
        }
    }
 
#if ARCH_X86
    if (   (uintptr_t)dst[0]&15 || (uintptr_t)dst[1]&15 || (uintptr_t)dst[2]&15
        || (uintptr_t)src[0]&15 || (uintptr_t)src[1]&15 || (uintptr_t)src[2]&15
        || dstStride[0]&15 || dstStride[1]&15 || dstStride[2]&15 || dstStride[3]&15
        || srcStride[0]&15 || srcStride[1]&15 || srcStride[2]&15 || srcStride[3]&15
    ) {
        SwsContext *const ctx = c->parent ? c->parent : c;
        int cpu_flags = av_get_cpu_flags();
        if (flags & SWS_PRINT_INFO && HAVE_MMXEXT && (cpu_flags & AV_CPU_FLAG_SSE2) &&
            !atomic_exchange_explicit(&ctx->stride_unaligned_warned,1, memory_order_relaxed)) {
            av_log(c, AV_LOG_WARNING, "Warning: data is not aligned! This can lead to a speed loss\n");
        }
    }
#endif
 
    if (scale_dst) {
        dstY         = dstSliceY;
        dstH         = dstY + dstSliceH;
        lastInLumBuf = -1;
        lastInChrBuf = -1;
    } else if (srcSliceY == 0) {
        /* Note the user might start scaling the picture in the middle so this
         * will not get executed. This is not really intended but works
         * currently, so people might do it. */
        dstY         = 0;
        lastInLumBuf = -1;
        lastInChrBuf = -1;
    }
 
    if (!should_dither) {
        c->chrDither8 = c->lumDither8 = sws_pb_64;
    }
    lastDstY = dstY;
 
    ff_init_vscale_pfn(c, yuv2plane1, yuv2planeX, yuv2nv12cX,
                   yuv2packed1, yuv2packed2, yuv2packedX, yuv2anyX, c->use_mmx_vfilter);
 
    ff_init_slice_from_src(src_slice, (uint8_t**)src, srcStride, c->srcW,
            srcSliceY, srcSliceH, chrSrcSliceY, chrSrcSliceH, 1);
 
    ff_init_slice_from_src(vout_slice, (uint8_t**)dst, dstStride, c->dstW,
            dstY, dstSliceH, dstY >> c->chrDstVSubSample,
            AV_CEIL_RSHIFT(dstSliceH, c->chrDstVSubSample), scale_dst);
    if (srcSliceY == 0) {
        hout_slice->plane[0].sliceY = lastInLumBuf + 1;
        hout_slice->plane[1].sliceY = lastInChrBuf + 1;
        hout_slice->plane[2].sliceY = lastInChrBuf + 1;
        hout_slice->plane[3].sliceY = lastInLumBuf + 1;
 
        hout_slice->plane[0].sliceH =
        hout_slice->plane[1].sliceH =
        hout_slice->plane[2].sliceH =
        hout_slice->plane[3].sliceH = 0;
        hout_slice->width = dstW;
    }
 
    for (; dstY < dstH; dstY++) {
        const int chrDstY = dstY >> c->chrDstVSubSample;
        int use_mmx_vfilter= c->use_mmx_vfilter;
 
        // First line needed as input
        const int firstLumSrcY  = FFMAX(1 - vLumFilterSize, vLumFilterPos[dstY]);
        const int firstLumSrcY2 = FFMAX(1 - vLumFilterSize, vLumFilterPos[FFMIN(dstY | ((1 << c->chrDstVSubSample) - 1), c->dstH - 1)]);
        // First line needed as input
        const int firstChrSrcY  = FFMAX(1 - vChrFilterSize, vChrFilterPos[chrDstY]);
 
        // Last line needed as input
        int lastLumSrcY  = FFMIN(c->srcH,    firstLumSrcY  + vLumFilterSize) - 1;
        int lastLumSrcY2 = FFMIN(c->srcH,    firstLumSrcY2 + vLumFilterSize) - 1;
        int lastChrSrcY  = FFMIN(c->chrSrcH, firstChrSrcY  + vChrFilterSize) - 1;
        int enough_lines;
 
        int i;
        int posY, cPosY, firstPosY, lastPosY, firstCPosY, lastCPosY;
 
        // handle holes (FAST_BILINEAR & weird filters)
        if (firstLumSrcY > lastInLumBuf) {
 
            hasLumHoles = lastInLumBuf != firstLumSrcY - 1;
            if (hasLumHoles) {
                hout_slice->plane[0].sliceY = firstLumSrcY;
                hout_slice->plane[3].sliceY = firstLumSrcY;
                hout_slice->plane[0].sliceH =
                hout_slice->plane[3].sliceH = 0;
            }
 
            lastInLumBuf = firstLumSrcY - 1;
        }
        if (firstChrSrcY > lastInChrBuf) {
 
            hasChrHoles = lastInChrBuf != firstChrSrcY - 1;
            if (hasChrHoles) {
                hout_slice->plane[1].sliceY = firstChrSrcY;
                hout_slice->plane[2].sliceY = firstChrSrcY;
                hout_slice->plane[1].sliceH =
                hout_slice->plane[2].sliceH = 0;
            }
 
            lastInChrBuf = firstChrSrcY - 1;
        }
 
        DEBUG_BUFFERS("dstY: %d\n", dstY);
        DEBUG_BUFFERS("\tfirstLumSrcY: %d lastLumSrcY: %d lastInLumBuf: %d\n",
                      firstLumSrcY, lastLumSrcY, lastInLumBuf);
        DEBUG_BUFFERS("\tfirstChrSrcY: %d lastChrSrcY: %d lastInChrBuf: %d\n",
                      firstChrSrcY, lastChrSrcY, lastInChrBuf);
 
        // Do we have enough lines in this slice to output the dstY line
        enough_lines = lastLumSrcY2 < srcSliceY + srcSliceH &&
                       lastChrSrcY < AV_CEIL_RSHIFT(srcSliceY + srcSliceH, c->chrSrcVSubSample);
 
        if (!enough_lines) {
            lastLumSrcY = srcSliceY + srcSliceH - 1;
            lastChrSrcY = chrSrcSliceY + chrSrcSliceH - 1;
            DEBUG_BUFFERS("buffering slice: lastLumSrcY %d lastChrSrcY %d\n",
                          lastLumSrcY, lastChrSrcY);
        }
 
        av_assert0((lastLumSrcY - firstLumSrcY + 1) <= hout_slice->plane[0].available_lines);
        av_assert0((lastChrSrcY - firstChrSrcY + 1) <= hout_slice->plane[1].available_lines);
 
 
        posY = hout_slice->plane[0].sliceY + hout_slice->plane[0].sliceH;
        if (posY <= lastLumSrcY && !hasLumHoles) {
            firstPosY = FFMAX(firstLumSrcY, posY);
            lastPosY = FFMIN(firstLumSrcY + hout_slice->plane[0].available_lines - 1, srcSliceY + srcSliceH - 1);
        } else {
            firstPosY = posY;
            lastPosY = lastLumSrcY;
        }
 
        cPosY = hout_slice->plane[1].sliceY + hout_slice->plane[1].sliceH;
        if (cPosY <= lastChrSrcY && !hasChrHoles) {
            firstCPosY = FFMAX(firstChrSrcY, cPosY);
            lastCPosY = FFMIN(firstChrSrcY + hout_slice->plane[1].available_lines - 1, AV_CEIL_RSHIFT(srcSliceY + srcSliceH, c->chrSrcVSubSample) - 1);
        } else {
            firstCPosY = cPosY;
            lastCPosY = lastChrSrcY;
        }
 
        ff_rotate_slice(hout_slice, lastPosY, lastCPosY);
 
        if (posY < lastLumSrcY + 1) {
            for (i = lumStart; i < lumEnd; ++i)
                desc[i].process(c, &desc[i], firstPosY, lastPosY - firstPosY + 1);
        }
 
        lastInLumBuf = lastLumSrcY;
 
        if (cPosY < lastChrSrcY + 1) {
            for (i = chrStart; i < chrEnd; ++i)
                desc[i].process(c, &desc[i], firstCPosY, lastCPosY - firstCPosY + 1);
        }
 
        lastInChrBuf = lastChrSrcY;
 
        if (!enough_lines)
            break;  // we can't output a dstY line so let's try with the next slice
 
#if HAVE_MMX_INLINE
        ff_updateMMXDitherTables(c, dstY);
#endif
        if (should_dither) {
            c->chrDither8 = ff_dither_8x8_128[chrDstY & 7];
            c->lumDither8 = ff_dither_8x8_128[dstY    & 7];
        }
        if (dstY >= c->dstH - 2) {
            /* hmm looks like we can't use MMX here without overwriting
             * this array's tail */
            ff_sws_init_output_funcs(c, &yuv2plane1, &yuv2planeX, &yuv2nv12cX,
                                     &yuv2packed1, &yuv2packed2, &yuv2packedX, &yuv2anyX);
            use_mmx_vfilter= 0;
            ff_init_vscale_pfn(c, yuv2plane1, yuv2planeX, yuv2nv12cX,
                           yuv2packed1, yuv2packed2, yuv2packedX, yuv2anyX, use_mmx_vfilter);
        }
 
        for (i = vStart; i < vEnd; ++i)
            desc[i].process(c, &desc[i], dstY, 1);
    }
    if (isPlanar(dstFormat) && isALPHA(dstFormat) && !needAlpha) {
        int offset = lastDstY - dstSliceY;
        int length = dstW;
        int height = dstY - lastDstY;
 
        if (is16BPS(dstFormat) || isNBPS(dstFormat)) {
            const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(dstFormat);
            fillPlane16(dst[3], dstStride[3], length, height, offset,
                    1, desc->comp[3].depth,
                    isBE(dstFormat));
        } else if (is32BPS(dstFormat)) {
            const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(dstFormat);
            fillPlane32(dst[3], dstStride[3], length, height, offset,
                    1, desc->comp[3].depth,
                    isBE(dstFormat), desc->flags & AV_PIX_FMT_FLAG_FLOAT);
        } else
            fillPlane(dst[3], dstStride[3], length, height, offset, 255);
    }
 
#if HAVE_MMXEXT_INLINE
    if (av_get_cpu_flags() & AV_CPU_FLAG_MMXEXT)
        __asm__ volatile ("sfence" ::: "memory");
#endif
    emms_c();
 
    /* store changed local vars back in the context */
    c->dstY         = dstY;
    c->lastInLumBuf = lastInLumBuf;
    c->lastInChrBuf = lastInChrBuf;
 
    return dstY - lastDstY;
}

可以看出swscale()是一行一行的进行图像缩放工作的。其中每行数据的处理按照“先水平拉伸，然后垂直拉伸”的方式进行处理。
具体的实现函数如下所示：
1. 水平拉伸
- a) 亮度水平拉伸：hyscale()
- b) 色度水平拉伸：hcscale()
2. 垂直拉伸
a) Planar
- i. 亮度垂直拉伸-不拉伸：yuv2plane1()
- ii. 亮度垂直拉伸-拉伸：yuv2planeX()
- iii. 色度垂直拉伸-不拉伸：yuv2plane1()
- iv. 色度垂直拉伸-拉伸：yuv2planeX()
b) Packed
- i. 垂直拉伸-不拉伸：yuv2packed1()
- ii. 垂直拉伸-拉伸：yuv2packedX()
下面具体看看这几个函数的定义。

hyscale()

水平亮度拉伸函数hyscale()的定义位于libswscale\swscale.c，如下所示。并不存在


    /**
     * Scale one horizontal line of input data using a filter over the input
     * lines, to produce one (differently sized) line of output data.
     *
     * @param dst        pointer to destination buffer for horizontally scaled
     *                   data. If the number of bits per component of one
     *                   destination pixel (SwsContext->dstBpc) is <= 10, data
     *                   will be 15 bpc in 16 bits (int16_t) width. Else (i.e.
     *                   SwsContext->dstBpc == 16), data will be 19bpc in
     *                   32 bits (int32_t) width.
     * @param dstW       width of destination image
     * @param src        pointer to source data to be scaled. If the number of
     *                   bits per component of a source pixel (SwsContext->srcBpc)
     *                   is 8, this is 8bpc in 8 bits (uint8_t) width. Else
     *                   (i.e. SwsContext->dstBpc > 8), this is native depth
     *                   in 16 bits (uint16_t) width. In other words, for 9-bit
     *                   YUV input, this is 9bpc, for 10-bit YUV input, this is
     *                   10bpc, and for 16-bit RGB or YUV, this is 16bpc.
     * @param filter     filter coefficients to be used per output pixel for
     *                   scaling. This contains 14bpp filtering coefficients.
     *                   Guaranteed to contain dstW * filterSize entries.
     * @param filterPos  position of the first input pixel to be used for
     *                   each output pixel during scaling. Guaranteed to
     *                   contain dstW entries.
     * @param filterSize the number of input coefficients to be used (and
     *                   thus the number of input pixels to be used) for
     *                   creating a single output pixel. Is aligned to 4
     *                   (and input coefficients thus padded with zeroes)
     *                   to simplify creating SIMD code.
     */
    /** @{ */
    void (*hyScale)(struct SwsContext *c, int16_t *dst, int dstW,
                    const uint8_t *src, const int16_t *filter,
                    const int32_t *filterPos, int filterSize);
    void (*hcScale)(struct SwsContext *c, int16_t *dst, int dstW,
                    const uint8_t *src, const int16_t *filter,
                    const int32_t *filterPos, int filterSize);

缺失内容

相关阅读:
字符串的方法
 通讯网关软件008——利用CommGate X2Mysql实现OPC数据转储Mysql
议程公布！Web3 建设者汇聚 DESTINATION MOON 分享见解与探讨
 基于C++和QT实现的二进制数独游戏求解
 Databend 开源周报第 114 期
 Apache Storm 2.5.0 集群安装与配置
 Linux 用户和组管理
 git仓库中增加子仓库
 php一句话木马免杀
 设计模式：单例模式（C#、JAVA、JavaScript、C++、Python、Go、PHP）
原文地址：https://blog.csdn.net/CHYabc123456hh/article/details/125434533