swscale: do not drop half of bits from 16bit bayer formats

4 files changed, 163 insertions, 2 deletions

diff --git a/libswscale/bayer_template.c b/libswscale/bayer_template.c
index 1af1b6056e..46b5a4984d 100644
--- a/libswscale/bayer_template.c
+++ b/libswscale/bayer_template.c
@@ -118,6 +118,72 @@
     B(1, 1) = (T(0, 1) + T(2, 1)) >> (1 + BAYER_SHIFT);
 #endif
 
+#if defined(BAYER_BGGR) || defined(BAYER_RGGB)
+#define BAYER_TO_RGB48_COPY \
+    R(0, 0) = \
+    R(0, 1) = \
+    R(1, 1) = \
+    R(1, 0) = S(1, 1); \
+    \
+    G(0, 1) = S(0, 1); \
+    G(0, 0) = \
+    G(1, 1) = (T(0, 1) + T(1, 0)) >> 1; \
+    G(1, 0) = S(1, 0); \
+    \
+    B(1, 1) = \
+    B(0, 0) = \
+    B(0, 1) = \
+    B(1, 0) = S(0, 0);
+#define BAYER_TO_RGB48_INTERPOLATE \
+    R(0, 0) = (T(-1, -1) + T(-1,  1) + T(1, -1) + T(1, 1)) >> 2; \
+    G(0, 0) = (T(-1,  0) + T( 0, -1) + T(0,  1) + T(1, 0)) >> 2; \
+    B(0, 0) =  S(0, 0); \
+    \
+    R(0, 1) = (T(-1, 1) + T(1, 1)) >> 1; \
+    G(0, 1) =  S(0,  1); \
+    B(0, 1) = (T(0,  0) + T(0, 2)) >> 1; \
+    \
+    R(1, 0) = (T(1, -1) + T(1, 1)) >> 1; \
+    G(1, 0) =  S(1,  0); \
+    B(1, 0) = (T(0,  0) + T(2, 0)) >> 1; \
+    \
+    R(1, 1) =  S(1, 1); \
+    G(1, 1) = (T(0, 1) + T(1, 0) + T(1, 2) + T(2, 1)) >> 2; \
+    B(1, 1) = (T(0, 0) + T(0, 2) + T(2, 0) + T(2, 2)) >> 2;
+#else
+#define BAYER_TO_RGB48_COPY \
+    R(0, 0) = \
+    R(0, 1) = \
+    R(1, 1) = \
+    R(1, 0) = S(1, 0); \
+    \
+    G(0, 0) = S(0, 0); \
+    G(1, 1) = S(1, 1); \
+    G(0, 1) = \
+    G(1, 0) = (T(0, 0) + T(1, 1)) >> 1; \
+    \
+    B(1, 1) = \
+    B(0, 0) = \
+    B(0, 1) = \
+    B(1, 0) = S(0, 1);
+#define BAYER_TO_RGB48_INTERPOLATE \
+    R(0, 0) = (T(-1, 0) + T(1, 0)) >> 1; \
+    G(0, 0) =  S(0, 0); \
+    B(0, 0) = (T(0, -1) + T(0, 1)) >> 1; \
+    \
+    R(0, 1) = (T(-1, 0) + T(-1, 2) + T(1, 0) + T(1, 2)) >> 2; \
+    G(0, 1) = (T(-1, 1) + T(0,  0) + T(0, 2) + T(1, 1)) >> 2; \
+    B(0, 1) =  S(0, 1); \
+    \
+    R(1, 0) =  S(1, 0); \
+    G(1, 0) = (T(0, 0)  + T(1, -1) + T(1,  1) + T(2, 0)) >> 2; \
+    B(1, 0) = (T(0, -1) + T(0,  1) + T(2, -1) + T(2, 1)) >> 2; \
+    \
+    R(1, 1) = (T(1, 0) + T(1, 2)) >> 1; \
+    G(1, 1) =  S(1, 1); \
+    B(1, 1) = (T(0, 1) + T(2, 1)) >> 1;
+#endif
+
 /**
  * invoke ff_rgb24toyv12 for 2x2 pixels
  */
@@ -153,6 +219,40 @@ static void BAYER_RENAME(rgb24_interpolate)(const uint8_t *src, int src_stride,
     }
 }
 
+static void BAYER_RENAME(rgb48_copy)(const uint8_t *src, int src_stride, uint8_t *ddst, int dst_stride, int width)
+{
+    uint16_t *dst = (uint16_t *)ddst;
+    int i;
+
+    dst_stride /= 2;
+    for (i = 0 ; i < width; i+= 2) {
+        BAYER_TO_RGB48_COPY
+        src += 2 * BAYER_SIZEOF;
+        dst += 6;
+    }
+}
+
+static void BAYER_RENAME(rgb48_interpolate)(const uint8_t *src, int src_stride, uint8_t *ddst, int dst_stride, int width)
+{
+    uint16_t *dst = (uint16_t *)ddst;
+    int i;
+
+    dst_stride /= 2;
+    BAYER_TO_RGB48_COPY
+    src += 2 * BAYER_SIZEOF;
+    dst += 6;
+
+    for (i = 2 ; i < width - 2; i+= 2) {
+        BAYER_TO_RGB48_INTERPOLATE
+        src += 2 * BAYER_SIZEOF;
+        dst += 6;
+    }
+
+    if (width > 2) {
+        BAYER_TO_RGB48_COPY
+    }
+}
+
 static void BAYER_RENAME(yv12_copy)(const uint8_t *src, int src_stride, uint8_t *dstY, uint8_t *dstU, uint8_t *dstV, int luma_stride, int width, int32_t *rgb2yuv)
 {
     uint8_t dst[12];
@@ -203,6 +303,8 @@ static void BAYER_RENAME(yv12_interpolate)(const uint8_t *src, int src_stride, u
 #undef B
 #undef BAYER_TO_RGB24_COPY
 #undef BAYER_TO_RGB24_INTERPOLATE
+#undef BAYER_TO_RGB48_COPY
+#undef BAYER_TO_RGB48_INTERPOLATE
 
 #undef BAYER_RENAME
 
diff --git a/libswscale/swscale_internal.h b/libswscale/swscale_internal.h
index 1a1b6f0dee..d207d3beff 100644
--- a/libswscale/swscale_internal.h
+++ b/libswscale/swscale_internal.h
@@ -764,6 +764,13 @@ static av_always_inline int isBayer(enum AVPixelFormat pix_fmt)
     return !!(desc->flags & AV_PIX_FMT_FLAG_BAYER);
 }
 
+static av_always_inline int isBayer16BPS(enum AVPixelFormat pix_fmt)
+{
+    const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
+    av_assert0(desc);
+    return desc->comp[1].depth == 8;
+}
+
 static av_always_inline int isAnyRGB(enum AVPixelFormat pix_fmt)
 {
     const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
diff --git a/libswscale/swscale_unscaled.c b/libswscale/swscale_unscaled.c
index 5fb572b51a..563de39696 100644
--- a/libswscale/swscale_unscaled.c
+++ b/libswscale/swscale_unscaled.c
@@ -1326,6 +1326,55 @@ static int bayer_to_rgb24_wrapper(SwsContext *c, const uint8_t* src[], int srcSt
     return srcSliceH;
 }
 
+static int bayer_to_rgb48_wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
+                                  int srcSliceH, uint8_t* dst[], int dstStride[])
+{
+    uint8_t *dstPtr= dst[0] + srcSliceY * dstStride[0];
+    const uint8_t *srcPtr= src[0];
+    int i;
+    void (*copy)       (const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride, int width);
+    void (*interpolate)(const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride, int width);
+
+    switch(c->srcFormat) {
+#define CASE(pixfmt, prefix) \
+    case pixfmt: copy        = bayer_##prefix##_to_rgb48_copy; \
+                 interpolate = bayer_##prefix##_to_rgb48_interpolate; \
+                 break;
+    CASE(AV_PIX_FMT_BAYER_BGGR8,    bggr8)
+    CASE(AV_PIX_FMT_BAYER_BGGR16LE, bggr16le)
+    CASE(AV_PIX_FMT_BAYER_BGGR16BE, bggr16be)
+    CASE(AV_PIX_FMT_BAYER_RGGB8,    rggb8)
+    CASE(AV_PIX_FMT_BAYER_RGGB16LE, rggb16le)
+    CASE(AV_PIX_FMT_BAYER_RGGB16BE, rggb16be)
+    CASE(AV_PIX_FMT_BAYER_GBRG8,    gbrg8)
+    CASE(AV_PIX_FMT_BAYER_GBRG16LE, gbrg16le)
+    CASE(AV_PIX_FMT_BAYER_GBRG16BE, gbrg16be)
+    CASE(AV_PIX_FMT_BAYER_GRBG8,    grbg8)
+    CASE(AV_PIX_FMT_BAYER_GRBG16LE, grbg16le)
+    CASE(AV_PIX_FMT_BAYER_GRBG16BE, grbg16be)
+#undef CASE
+    default: return 0;
+    }
+
+    av_assert0(srcSliceH > 1);
+
+    copy(srcPtr, srcStride[0], dstPtr, dstStride[0], c->srcW);
+    srcPtr += 2 * srcStride[0];
+    dstPtr += 2 * dstStride[0];
+
+    for (i = 2; i < srcSliceH - 2; i += 2) {
+        interpolate(srcPtr, srcStride[0], dstPtr, dstStride[0], c->srcW);
+        srcPtr += 2 * srcStride[0];
+        dstPtr += 2 * dstStride[0];
+    }
+
+    if (i + 1 == srcSliceH) {
+        copy(srcPtr, -srcStride[0], dstPtr, -dstStride[0], c->srcW);
+    } else if (i < srcSliceH)
+        copy(srcPtr, srcStride[0], dstPtr, dstStride[0], c->srcW);
+    return srcSliceH;
+}
+
 static int bayer_to_yv12_wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
                                  int srcSliceH, uint8_t* dst[], int dstStride[])
 {
@@ -2047,6 +2096,8 @@ void ff_get_unscaled_swscale(SwsContext *c)
     if (isBayer(srcFormat)) {
         if (dstFormat == AV_PIX_FMT_RGB24)
             c->swscale = bayer_to_rgb24_wrapper;
+        else if (dstFormat == AV_PIX_FMT_RGB48)
+            c->swscale = bayer_to_rgb48_wrapper;
         else if (dstFormat == AV_PIX_FMT_YUV420P)
             c->swscale = bayer_to_yv12_wrapper;
         else if (!isBayer(dstFormat)) {
diff --git a/libswscale/utils.c b/libswscale/utils.c
index dcd1dbaa76..9ca378bd3b 100644
--- a/libswscale/utils.c
+++ b/libswscale/utils.c
@@ -1536,8 +1536,9 @@ av_cold int sws_init_context(SwsContext *c, SwsFilter *srcFilter,
 
     if (isBayer(srcFormat)) {
         if (!unscaled ||
-            (dstFormat != AV_PIX_FMT_RGB24 && dstFormat != AV_PIX_FMT_YUV420P)) {
-            enum AVPixelFormat tmpFormat = AV_PIX_FMT_RGB24;
+            (dstFormat != AV_PIX_FMT_RGB24 && dstFormat != AV_PIX_FMT_YUV420P &&
+             dstFormat != AV_PIX_FMT_RGB48)) {
+            enum AVPixelFormat tmpFormat = isBayer16BPS(srcFormat) ? AV_PIX_FMT_RGB48 : AV_PIX_FMT_RGB24;
 
             ret = av_image_alloc(c->cascaded_tmp, c->cascaded_tmpStride,
                                 srcW, srcH, tmpFormat, 64);