A novel integration of image compression and sensing is proposed to enhance the performance of a CMOS image sensor. By integrating a compression function onto the sensor focal plane, the image signal to be read out can be significantly reduced and consequently the pixel rate can be increased. This can be applied to overcome the communication bottleneck for high-resolution image sensing and high frame-rate image sensing or for power- and bandwidth-constrained devices such as cell phones. A modified Haar wavelet transform is implemented as the compression scheme. A simple but efficient computation design is developed to implement the transform on-chip.