Convolution is a basic signal and image processing ap- plication. In image processing, kernel coefficients of con- volution commonly remain constant across the entire im- age. A less common situation is where the kernel coeffi- cients change in value for each pixel in the image. We call this variable kernel convolution. In this paper we present what we believe are the first three FPGA implementations of variable kernel convolution. The first uses sequential streaming, the second uses pipelining and the third solu- tion uses what we call convolve and gather and its hardware implementation has the highest area time rating (6.7 ? bet- ter than streaming and 3.4 ? better than the pipelining so- lution). Both pipelining and convolve and gather have the same throughput (which is 25? that of streaming), but con- volve and gather has 71% smaller area footprint than the pipeline.