diff --git a/src/ImageSharp/Processing/Processors/Convolution/Convolution2DProcessor{TPixel}.cs b/src/ImageSharp/Processing/Processors/Convolution/Convolution2DProcessor{TPixel}.cs
index b8a61e5a32..8a7c424815 100644
--- a/src/ImageSharp/Processing/Processors/Convolution/Convolution2DProcessor{TPixel}.cs
+++ b/src/ImageSharp/Processing/Processors/Convolution/Convolution2DProcessor{TPixel}.cs
@@ -64,10 +64,6 @@ internal class Convolution2DProcessor<TPixel> : ImageProcessor<TPixel>
 
         var interest = Rectangle.Intersect(this.SourceRectangle, source.Bounds());
 
-        // We use a rectangle 3x the interest width to allocate a buffer big enough
-        // for source and target bulk pixel conversion.
-        var operationBounds = new Rectangle(interest.X, interest.Y, interest.Width * 3, interest.Height);
-
         using (var map = new KernelSamplingMap(allocator))
         {
             // Since the kernel sizes are identical we can use a single map.
@@ -85,7 +81,7 @@ internal class Convolution2DProcessor<TPixel> : ImageProcessor<TPixel>
 
             ParallelRowIterator.IterateRows<Convolution2DRowOperation<TPixel>, Vector4>(
                 this.Configuration,
-                operationBounds,
+                interest,
                 in operation);
         }
 
diff --git a/src/ImageSharp/Processing/Processors/Convolution/Convolution2DRowOperation{TPixel}.cs b/src/ImageSharp/Processing/Processors/Convolution/Convolution2DRowOperation{TPixel}.cs
index c04581444f..e5963bd390 100644
--- a/src/ImageSharp/Processing/Processors/Convolution/Convolution2DRowOperation{TPixel}.cs
+++ b/src/ImageSharp/Processing/Processors/Convolution/Convolution2DRowOperation{TPixel}.cs
@@ -49,7 +49,7 @@ internal readonly struct Convolution2DRowOperation<TPixel> : IRowOperation<Vecto
     /// <inheritdoc/>
     [MethodImpl(InliningOptions.ShortMethod)]
     public int GetRequiredBufferLength(Rectangle bounds)
-        => bounds.Width;
+        => 3 * bounds.Width;
 
     /// <inheritdoc/>
     [MethodImpl(MethodImplOptions.AggressiveInlining)]
diff --git a/src/ImageSharp/Processing/Processors/Convolution/Convolution2PassProcessor{TPixel}.cs b/src/ImageSharp/Processing/Processors/Convolution/Convolution2PassProcessor{TPixel}.cs
index fa05a98fd1..094a96f787 100644
--- a/src/ImageSharp/Processing/Processors/Convolution/Convolution2PassProcessor{TPixel}.cs
+++ b/src/ImageSharp/Processing/Processors/Convolution/Convolution2PassProcessor{TPixel}.cs
@@ -70,10 +70,6 @@ internal class Convolution2PassProcessor<TPixel> : ImageProcessor<TPixel>
 
         var interest = Rectangle.Intersect(this.SourceRectangle, source.Bounds());
 
-        // We use a rectangle 2x the interest width to allocate a buffer big enough
-        // for source and target bulk pixel conversion.
-        var operationBounds = new Rectangle(interest.X, interest.Y, interest.Width * 2, interest.Height);
-
         // We can create a single sampling map with the size as if we were using the non separated 2D kernel
         // the two 1D kernels represent, and reuse it across both convolution steps, like in the bokeh blur.
         using var mapXY = new KernelSamplingMap(this.Configuration.MemoryAllocator);
@@ -92,7 +88,7 @@ internal class Convolution2PassProcessor<TPixel> : ImageProcessor<TPixel>
 
         ParallelRowIterator.IterateRows<HorizontalConvolutionRowOperation, Vector4>(
             this.Configuration,
-            operationBounds,
+            interest,
             in horizontalOperation);
 
         // Vertical convolution
@@ -107,7 +103,7 @@ internal class Convolution2PassProcessor<TPixel> : ImageProcessor<TPixel>
 
         ParallelRowIterator.IterateRows<VerticalConvolutionRowOperation, Vector4>(
             this.Configuration,
-            operationBounds,
+            interest,
             in verticalOperation);
     }
 
@@ -146,7 +142,7 @@ internal class Convolution2PassProcessor<TPixel> : ImageProcessor<TPixel>
         /// <inheritdoc/>
         [MethodImpl(InliningOptions.ShortMethod)]
         public int GetRequiredBufferLength(Rectangle bounds)
-            => bounds.Width;
+            => 2 * bounds.Width;
 
         /// <inheritdoc/>
         [MethodImpl(MethodImplOptions.AggressiveInlining)]
@@ -312,7 +308,7 @@ internal class Convolution2PassProcessor<TPixel> : ImageProcessor<TPixel>
         /// <inheritdoc/>
         [MethodImpl(InliningOptions.ShortMethod)]
         public int GetRequiredBufferLength(Rectangle bounds)
-            => bounds.Width;
+            => 2 * bounds.Width;
 
         /// <inheritdoc/>
         [MethodImpl(MethodImplOptions.AggressiveInlining)]
diff --git a/src/ImageSharp/Processing/Processors/Convolution/ConvolutionProcessor{TPixel}.cs b/src/ImageSharp/Processing/Processors/Convolution/ConvolutionProcessor{TPixel}.cs
index f20cca3f71..54dad64a6b 100644
--- a/src/ImageSharp/Processing/Processors/Convolution/ConvolutionProcessor{TPixel}.cs
+++ b/src/ImageSharp/Processing/Processors/Convolution/ConvolutionProcessor{TPixel}.cs
@@ -57,9 +57,6 @@ internal class ConvolutionProcessor<TPixel> : ImageProcessor<TPixel>
 
         var interest = Rectangle.Intersect(this.SourceRectangle, source.Bounds());
 
-        // We use a rectangle 2x the interest width to allocate a buffer big enough
-        // for source and target bulk pixel conversion.
-        var operationBounds = new Rectangle(interest.X, interest.Y, interest.Width * 2, interest.Height);
         using (var map = new KernelSamplingMap(allocator))
         {
             map.BuildSamplingOffsetMap(this.KernelXY, interest);
@@ -67,7 +64,7 @@ internal class ConvolutionProcessor<TPixel> : ImageProcessor<TPixel>
             var operation = new RowOperation(interest, targetPixels, source.PixelBuffer, map, this.KernelXY, this.Configuration, this.PreserveAlpha);
             ParallelRowIterator.IterateRows<RowOperation, Vector4>(
                this.Configuration,
-               operationBounds,
+               interest,
                in operation);
         }
 
@@ -109,7 +106,7 @@ internal class ConvolutionProcessor<TPixel> : ImageProcessor<TPixel>
         /// <inheritdoc/>
         [MethodImpl(InliningOptions.ShortMethod)]
         public int GetRequiredBufferLength(Rectangle bounds)
-            => bounds.Width;
+            => 2 * bounds.Width;
 
         /// <inheritdoc/>
         [MethodImpl(InliningOptions.ShortMethod)]
diff --git a/src/ImageSharp/Processing/Processors/Convolution/MedianBlurProcessor{TPixel}.cs b/src/ImageSharp/Processing/Processors/Convolution/MedianBlurProcessor{TPixel}.cs
index 4f0c2a36c9..fe3a29d437 100644
--- a/src/ImageSharp/Processing/Processors/Convolution/MedianBlurProcessor{TPixel}.cs
+++ b/src/ImageSharp/Processing/Processors/Convolution/MedianBlurProcessor{TPixel}.cs
@@ -31,11 +31,6 @@ internal sealed class MedianBlurProcessor<TPixel> : ImageProcessor<TPixel>
 
         Rectangle interest = Rectangle.Intersect(this.SourceRectangle, source.Bounds());
 
-        // We use a rectangle with width set wider, to allocate a buffer big enough
-        // for kernel source, channel buffers, source rows and target bulk pixel conversion.
-        int operationWidth = (2 * kernelSize * kernelSize) + interest.Width + (kernelSize * interest.Width);
-        Rectangle operationBounds = new(interest.X, interest.Y, operationWidth, interest.Height);
-
         using KernelSamplingMap map = new(this.Configuration.MemoryAllocator);
         map.BuildSamplingOffsetMap(kernelSize, kernelSize, interest, this.definition.BorderWrapModeX, this.definition.BorderWrapModeY);
 
@@ -50,7 +45,7 @@ internal sealed class MedianBlurProcessor<TPixel> : ImageProcessor<TPixel>
 
         ParallelRowIterator.IterateRows<MedianRowOperation<TPixel>, Vector4>(
             this.Configuration,
-            operationBounds,
+            interest,
             in operation);
 
         Buffer2D<TPixel>.SwapOrCopyContent(source.PixelBuffer, targetPixels);
diff --git a/src/ImageSharp/Processing/Processors/Convolution/MedianRowOperation{TPixel}.cs b/src/ImageSharp/Processing/Processors/Convolution/MedianRowOperation{TPixel}.cs
index 17df7b9be4..8caf33440d 100644
--- a/src/ImageSharp/Processing/Processors/Convolution/MedianRowOperation{TPixel}.cs
+++ b/src/ImageSharp/Processing/Processors/Convolution/MedianRowOperation{TPixel}.cs
@@ -46,11 +46,11 @@ internal readonly struct MedianRowOperation<TPixel> : IRowOperation<Vector4>
     /// <inheritdoc/>
     [MethodImpl(InliningOptions.ShortMethod)]
     public int GetRequiredBufferLength(Rectangle bounds)
-        => bounds.Width;
+        => (2 * this.kernelSize * this.kernelSize) + bounds.Width + (kernelSize * bounds.Width);
 
     public void Invoke(int y, Span<Vector4> span)
     {
-        // Span has kernelSize^2 followed by bound width.
+        // Span has kernelSize^2 twice, then bound width followed by kernelsize * bounds width.
         int boundsX = this.bounds.X;
         int boundsWidth = this.bounds.Width;
         int kernelCount = this.kernelSize * this.kernelSize;