cv::cvtColor (Lab2BGR CV_8U)

modules/imgproc/src/color.cpp

@@ -3773,7 +3773,14 @@ struct Lab2RGB_b
 
     Lab2RGB_b( int _dstcn, int blueIdx, const float* _coeffs,
                const float* _whitept, bool _srgb )
-    : dstcn(_dstcn), cvt(3, blueIdx, _coeffs, _whitept, _srgb ) {}
+    : dstcn(_dstcn), cvt(3, blueIdx, _coeffs, _whitept, _srgb )
+    {
+        #if CV_NEON
+        v_scale_inv = vdupq_n_f32(100.f/255.f);
+        v_scale = vdupq_n_f32(255.f);
+        v_alpha = vdup_n_u8(ColorChannel<uchar>::max());
+        #endif
+    }
 
     void operator()(const uchar* src, uchar* dst, int n) const
     {
@@ -3784,16 +3791,70 @@ struct Lab2RGB_b
         for( i = 0; i < n; i += BLOCK_SIZE, src += BLOCK_SIZE*3 )
         {
             int dn = std::min(n - i, (int)BLOCK_SIZE);
+            j = 0;
 
-            for( j = 0; j < dn*3; j += 3 )
+            #if CV_NEON
+            for ( ; j <= (dn - 8) * 3; j += 24)
+            {
+                uint8x8x3_t v_src = vld3_u8(src + j);
+                uint16x8_t v_t0 = vmovl_u8(v_src.val[0]),
+                           v_t1 = vmovl_u8(v_src.val[1]),
+                           v_t2 = vmovl_u8(v_src.val[2]);
+
+                float32x4x3_t v_dst;
+                v_dst.val[0] = vmulq_f32(vcvtq_f32_u32(vmovl_u16(vget_low_u16(v_t0))), v_scale_inv);
+                v_dst.val[1] = vsubq_f32(vcvtq_f32_u32(vmovl_u16(vget_low_u16(v_t1))), v_128);
+                v_dst.val[2] = vsubq_f32(vcvtq_f32_u32(vmovl_u16(vget_low_u16(v_t2))), v_128);
+                vst3q_f32(buf + j, v_dst);
+
+                v_dst.val[0] = vmulq_f32(vcvtq_f32_u32(vmovl_u16(vget_high_u16(v_t0))), v_scale_inv);
+                v_dst.val[1] = vsubq_f32(vcvtq_f32_u32(vmovl_u16(vget_high_u16(v_t1))), v_128);
+                v_dst.val[2] = vsubq_f32(vcvtq_f32_u32(vmovl_u16(vget_high_u16(v_t2))), v_128);
+                vst3q_f32(buf + j + 12, v_dst);
+            }
+            #endif
+
+            for( ; j < dn*3; j += 3 )
             {
                 buf[j] = src[j]*(100.f/255.f);
                 buf[j+1] = (float)(src[j+1] - 128);
                 buf[j+2] = (float)(src[j+2] - 128);
             }
             cvt(buf, buf, dn);
+            j = 0;
 
-            for( j = 0; j < dn*3; j += 3, dst += dcn )
+            #if CV_NEON
+            for ( ; j <= (dn - 8) * 3; j += 24, dst += dcn * 8)
+            {
+                float32x4x3_t v_src0 = vld3q_f32(buf + j), v_src1 = vld3q_f32(buf + j + 12);
+                uint8x8_t v_dst0 = vqmovn_u16(vcombine_u16(vqmovn_u32(cv_vrndq_u32_f32(vmulq_f32(v_src0.val[0], v_scale))),
+                                                           vqmovn_u32(cv_vrndq_u32_f32(vmulq_f32(v_src1.val[0], v_scale)))));
+                uint8x8_t v_dst1 = vqmovn_u16(vcombine_u16(vqmovn_u32(cv_vrndq_u32_f32(vmulq_f32(v_src0.val[1], v_scale))),
+                                                           vqmovn_u32(cv_vrndq_u32_f32(vmulq_f32(v_src1.val[1], v_scale)))));
+                uint8x8_t v_dst2 = vqmovn_u16(vcombine_u16(vqmovn_u32(cv_vrndq_u32_f32(vmulq_f32(v_src0.val[2], v_scale))),
+                                                           vqmovn_u32(cv_vrndq_u32_f32(vmulq_f32(v_src1.val[2], v_scale)))));
+
+                if (dcn == 4)
+                {
+                    uint8x8x4_t v_dst;
+                    v_dst.val[0] = v_dst0;
+                    v_dst.val[1] = v_dst1;
+                    v_dst.val[2] = v_dst2;
+                    v_dst.val[3] = v_alpha;
+                    vst4_u8(dst, v_dst);
+                }
+                else
+                {
+                    uint8x8x3_t v_dst;
+                    v_dst.val[0] = v_dst0;
+                    v_dst.val[1] = v_dst1;
+                    v_dst.val[2] = v_dst2;
+                    vst3_u8(dst, v_dst);
+                }
+            }
+            #endif
+
+            for( ; j < dn*3; j += 3, dst += dcn )
             {
                 dst[0] = saturate_cast<uchar>(buf[j]*255.f);
                 dst[1] = saturate_cast<uchar>(buf[j+1]*255.f);
@@ -3806,6 +3867,11 @@ struct Lab2RGB_b
 
     int dstcn;
     Lab2RGB_f cvt;
+
+    #if CV_NEON
+    float32x4_t v_scale, v_scale_inv, v_128;
+    uint8x8_t v_alpha;
+    #endif
 };