codec_support/blockdsp: add halfpel interpolation functions

[nihav.git] / nihav-codec-support / src / codecs / blockdsp.rs

diff --git a/nihav-codec-support/src/codecs/blockdsp.rs b/nihav-codec-support/src/codecs/blockdsp.rs
index da92742cec741a09514379aa029ea6eb4b5bf163..9f70582c17a09fbf4db08353e4a70f7bc46b9262 100644 (file)
--- a/nihav-codec-support/src/codecs/blockdsp.rs
+++ b/nihav-codec-support/src/codecs/blockdsp.rs
@@ -112,13 +112,22 @@ pub fn add_blocks(buf: &mut NAVideoBuffer<u8>, xpos: usize, ypos: usize, blk: &[
 }
 
 /// Copies block from the picture with pixels beyond the picture borders being replaced with replicated edge pixels.
-pub fn edge_emu(src: &NAVideoBuffer<u8>, xpos: isize, ypos: isize, bw: usize, bh: usize, dst: &mut [u8], dstride: usize, comp: usize) {
+pub fn edge_emu(src: &NAVideoBuffer<u8>, xpos: isize, ypos: isize, bw: usize, bh: usize, dst: &mut [u8], dstride: usize, comp: usize, align: u8) {
     let stride = src.get_stride(comp);
     let offs   = src.get_offset(comp);
-    let (w, h) = src.get_dimensions(comp);
+    let (w_, h_) = src.get_dimensions(comp);
+    let (hss, vss) = src.get_info().get_format().get_chromaton(comp).unwrap().get_subsampling();
     let data = src.get_data();
     let framebuf: &[u8] = data.as_slice();
 
+    let (w, h) = if align == 0 {
+            (w_, h_)
+        } else {
+            let wa = if align > hss { (1 << (align - hss)) - 1 } else { 0 };
+            let ha = if align > vss { (1 << (align - vss)) - 1 } else { 0 };
+            ((w_ + wa) & !wa, (h_ + ha) & !ha)
+        };
+
     for y in 0..bh {
         let srcy;
         if (y as isize) + ypos < 0 { srcy = 0; }
@@ -182,7 +191,7 @@ pub fn copy_blocks(dst: &mut NAVideoBuffer<u8>, src: &NAVideoBuffer<u8>,
             let bw_ = (if comp > 0 { bw/2 } else { bw }) + ((pre + post) as usize);
             let bh_ = (if comp > 0 { bh/2 } else { bh }) + ((pre + post) as usize);
             edge_emu(src, sx_ - pre, sy_ - pre, bw_, bh_,
-                     ebuf.as_mut_slice(), ebuf_stride, comp);
+                     ebuf.as_mut_slice(), ebuf_stride, comp, 0);
             let bw_ = if comp > 0 { bw/2 } else { bw };
             let bh_ = if comp > 0 { bh/2 } else { bh };
             (interp[mode])(&mut dbuf[doff + x + y * dstride..], dstride, ebuf.as_slice(), ebuf_stride, bw_, bh_);
@@ -233,7 +242,7 @@ pub fn copy_block(dst: &mut NASimpleVideoFrame<u8>, src: NAVideoBufferRef<u8>, c
         let doff    = dst.offset[comp];
         let edge = (pre + post) as usize;
         edge_emu(&src, sx - pre, sy - pre, bw + edge, bh + edge,
-                 ebuf.as_mut_slice(), ebuf_stride, comp);
+                 ebuf.as_mut_slice(), ebuf_stride, comp, 0);
         (interp[mode])(&mut dst.data[doff + dx + dy * dstride..], dstride,
                        ebuf.as_slice(), ebuf_stride, bw, bh);
     } else {
@@ -248,3 +257,57 @@ pub fn copy_block(dst: &mut NASimpleVideoFrame<u8>, src: NAVideoBufferRef<u8>, c
                        &sbuf[saddr..], sstride, bw, bh);
     }
 }
+
+fn hpel_interp00(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, bw: usize, bh: usize)
+{
+    let mut didx = 0;
+    let mut sidx = 0;
+    for _ in 0..bh {
+        dst[didx..][..bw].copy_from_slice(&src[sidx..][..bw]);
+        didx += dstride;
+        sidx += sstride;
+    }
+}
+
+fn hpel_interp01(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, bw: usize, bh: usize)
+{
+    let mut didx = 0;
+    let mut sidx = 0;
+    for _ in 0..bh {
+        for x in 0..bw { dst[didx + x] = ((u16::from(src[sidx + x]) + u16::from(src[sidx + x + 1]) + 1) >> 1) as u8; }
+        didx += dstride;
+        sidx += sstride;
+    }
+}
+
+fn hpel_interp10(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, bw: usize, bh: usize)
+{
+    let mut didx = 0;
+    let mut sidx = 0;
+    for _ in 0..bh {
+        for x in 0..bw { dst[didx + x] = ((u16::from(src[sidx + x]) + u16::from(src[sidx + x + sstride]) + 1) >> 1) as u8; }
+        didx += dstride;
+        sidx += sstride;
+    }
+}
+
+fn hpel_interp11(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, bw: usize, bh: usize)
+{
+    let mut didx = 0;
+    let mut sidx = 0;
+    for _ in 0..bh {
+        for x in 0..bw {
+            dst[didx + x] = ((u16::from(src[sidx + x]) +
+                              u16::from(src[sidx + x + 1]) +
+                              u16::from(src[sidx + x + sstride]) +
+                              u16::from(src[sidx + x + sstride + 1]) + 2) >> 2) as u8;
+        }
+        didx += dstride;
+        sidx += sstride;
+    }
+}
+
+/// Half-pixel interpolation functions.
+pub const HALFPEL_INTERP_FUNCS: &[BlkInterpFunc] = &[
+        hpel_interp00, hpel_interp01, hpel_interp10, hpel_interp11 ];
+