diff --git a/README.md b/README.md
index b073ef2..8635c03 100644
--- a/README.md
+++ b/README.md
@@ -229,7 +229,7 @@ FUNCTION     | NOTES
 [p_max()](src/math/p_max.c)           | finds max val  
 [p_min()](src/math/p_min.c)           | finds min val  
 [p_mean()](src/math/p_mean.c)         | mean operation  
-[p_median()](src/math/p_mean.c)       | finds middle value  
+[p_median()](src/math/p_median.c)     | finds middle value  
 [p_mode()](src/math/p_mode.c)         | finds most common value  
 [p_mul()](src/math/p_mul.c)           | multiplication  
 [p_popcount()](src/math/p_popcount.c) | count the number of bits set  
diff --git a/include/pal_image.h b/include/pal_image.h
index 6965542..9007b36 100644
--- a/include/pal_image.h
+++ b/include/pal_image.h
@@ -17,7 +17,7 @@
 
 /*2d convolution */
 void p_conv2d_f32(const float *x, float *m, float *r, int rows, int cols,
-                  int msize);
+                  int mrows, int mcols);
 
 /*2d box (i.e mean) filter(3x3) */
 void p_box3x3_f32(const float *x, float *r, int rows, int cols);
diff --git a/src/dsp/p_conv.c b/src/dsp/p_conv.c
index 60f2a20..acde422 100644
--- a/src/dsp/p_conv.c
+++ b/src/dsp/p_conv.c
@@ -22,11 +22,15 @@ void p_conv_f32(const float *x, const float *h, float *r, int nx, int nh)
 {
     const float *xc = x;
     float *rx = r;
-  	for ( int i = 0; i < nx; i++) {
+    int i,j ;
+        for ( i = 0; i < nx+nh-1; i++) { *(rx++) = 0; } 
+        rx = r ;
+  	for ( i = 0; i < nx; i++) {
         float xv = *xc++;
-        rx++;
-  		for (int j = 0; j < nh; j++) {
-  			*(rx + j) += xv * *(h + j);		
+
+  		for (j = 0; j < nh; j++) {
+  			*(rx + j) += xv * *(h + j);	
   		}
+        rx++;
   	}
 }
diff --git a/src/image/p_conv2d.c b/src/image/p_conv2d.c
index bff439b..808ae5a 100644
--- a/src/image/p_conv2d.c
+++ b/src/image/p_conv2d.c
@@ -12,43 +12,39 @@
  *
  * @param cols  Number of columns in input image
  *
- * @param msize Size of convolution kernel
+ * @param mrows number of rows in convolution kernel
+ *
+ * @param mcols number of cols in convolution kernel
  *
  * @return      None
  *
  */
 
 void p_conv2d_f32(const float *x, float *m, float *r, int rows, int cols,
-                  int msize)
+                  int mrows, int mcols)
 
 {
-    int i, j, k;
-    float P, part;
+    int i, j, ki, kj;
+    float P;
     const float *px, *pm;
     float *pr;
 
     px = x;
-    pm = m;
     pr = r;
 
-    for (i = msize * 0.5; i < (rows - msize * 0.5); i++) {
-        for (j = msize * 0.5; j < (cols - msize * 0.5); j++) {
+    for (i = 0; i < rows - mrows+1  ; i++) {
+        for (j = 0; j < cols - mcols+1 ; j++) {
             P = 0.0f;
-            pm = m;
-            for (k = 0; k < msize; k++) {
-                p_dot_f32(px, pm, &part, msize);
-                P += part;
-                px += cols;
-                pm += msize;
+            pm = m+(mcols * mrows)-1;
+            for (ki=0 ; ki< mrows ; ki++){
+                for (kj=0 ; kj< mcols ; kj++){
+                    P+= (*px++)* (*pm--) ;
+		}
+            px += cols - mcols;
             }
-            *pr = P;
-            pr++;
-            // move image pointer one index forward compared to
-            // the position from before `for` loop
-            px += 1 - msize * cols;
+            px -= (mrows * cols) -1 ;
+            *(pr++) = P;
         }
-        // move image pointer to the beginning of line
-        // beneath the current line
-        px += (int)(msize * 0.5) * 2;
+    px+=mcols-1 ;
     }
 }
diff --git a/src/image/p_median3x3.c b/src/image/p_median3x3.c
index b86303f..9210c01 100644
--- a/src/image/p_median3x3.c
+++ b/src/image/p_median3x3.c
@@ -1,5 +1,43 @@
 #include <pal.h>
+/*
+ * The following routines have been built from knowledge gathered
+ * around the Web. I am not aware of any copyright problem with
+ * them, so use it as you want.
+ * N. Devillard - 1998
+ */
+
+typedef float pixelvalue ;
+
+#define PIX_SORT(a,b) { if ((a)>(b)) PIX_SWAP((a),(b)); }
+#define PIX_SWAP(a,b) { pixelvalue temp=(a);(a)=(b);(b)=temp; }
 
+/*----------------------------------------------------------------------------
+   Function :   opt_med9()
+   In       :   pointer to an array of 9 pixelvalues
+   Out      :   a pixelvalue
+   Job      :   optimized search of the median of 9 pixelvalues
+   Notice   :   in theory, cannot go faster without assumptions on the
+                signal.
+                Formula from:
+                XILINX XCELL magazine, vol. 23 by John L. Smith
+  
+                The input array is *NOT* modified in the process
+                The result array is guaranteed to contain the median
+                value
+ ---------------------------------------------------------------------------*/
+
+pixelvalue opt_med9(pixelvalue * pointer)
+{
+    pixelvalue p[9];
+    memcpy(p, pointer, 9*sizeof(pixelvalue) );
+    PIX_SORT(p[1], p[2]) ; PIX_SORT(p[4], p[5]) ; PIX_SORT(p[7], p[8]) ;
+    PIX_SORT(p[0], p[1]) ; PIX_SORT(p[3], p[4]) ; PIX_SORT(p[6], p[7]) ;
+    PIX_SORT(p[1], p[2]) ; PIX_SORT(p[4], p[5]) ; PIX_SORT(p[7], p[8]) ;
+    PIX_SORT(p[0], p[3]) ; PIX_SORT(p[5], p[8]) ; PIX_SORT(p[4], p[7]) ;
+    PIX_SORT(p[3], p[6]) ; PIX_SORT(p[1], p[4]) ; PIX_SORT(p[2], p[5]) ;
+    PIX_SORT(p[4], p[7]) ; PIX_SORT(p[4], p[2]) ; PIX_SORT(p[6], p[4]) ;
+    PIX_SORT(p[4], p[2]) ; return(p[4]) ;
+}
 /*
  * A median 3x3 filter.
  *
@@ -50,7 +88,7 @@ void p_median3x3_f32(const float *x, float *r, int rows, int cols)
             buffer[buffer_col + 3] = *(px + cols);
             buffer[buffer_col + 6] = *(px + cols + cols);
 
-            p_median_f32(buffer, pr, 9);
+            *pr = opt_med9(buffer);
             pr++;
             px++;
         }