Image and video compression on resource-limited platforms like drones and robots is challenging due to the need for low power consumption and computational complexity. Existing codecs like MPEG are not suitable due to their high encoding complexity. In this work, the authors propose an ultra low-complexity image and video codec using block modulation, dubbed BMVC representing Block Modulating Video Compression codec. The underlying principle of BMVC is to mask the high resolution image (via pre-defined binary random coding patterns composed of {0,1}) and then decompose it into different small (modulated) blocks. Finally, these blocks are summed up to a single block and quantized as the compressed signal to be transmitted. Since no multiplication is involved during this encoding process, the complexity of the BMVC encoder is way lower than MPEG-based codecs. Moreover, the summation over modulated image blocks by binary masks can be essentially implemented as additions of pixel readouts according to a pre-defined look-up table during this encoding process. The results are published entitled "Block modulating video compression: an ultra low complexity image compression encoder for resource limited platforms" in Advanced Imaging.

Pipeline of the proposed Block Modulating Video Compression (BMVC) encoder.

The authors develop a Block Modulating Video Compression (BMVC) encoder that masks the image with a pre-defined binary mask and then sums the modulated blocks to obtain a single compressed block. This process only involves addition operations, significantly reducing the computational complexity. Besides, two decoders are also developed:

• BMVC-PnP: An iterative decoder based on the Plug-and-Play (PnP) algorithm, utilizing a pre-trained denoising network for regularization.
• BMVC-E2E: A feed-forward decoder based on an end-to-end convolutional neural network (CNN) architecture for real-time decoding.

Proposed Block Modulating Video Compression (BMVC) decoder.

The proposed method achieved high compression ratios (up to 150) while maintaining acceptable decoding quality, and significantly lower computational complexity compared to existing codecs like JPEG2000. Both BMVC-PnP and BMVC-E2E decoders demonstrated robustness against bit quantization, providing consistent performance across different quantization levels.

Overall, this work offers a promising solution for image and video compression on resource-limited platforms. Its ultra low-complexity encoder and high-quality decoders make it suitable for applications like drones and robots, where power and computational resources are limited.