Infinite Algorithms

Video is a three dimensional signal that is almost magical. Any thing that we can see can be converted to a video signal by placing a camera and capturing the light falling into the camera sensor. Then this signal can be processed, compressed, converted to binary streams and sent to any place you can imagine. In this way, one can be instantly be transported to another place and time. Conversely by having multiple video streams from different places, one can even "exist" in multiple places simultaneously!

Video is also a very complex entity to model and process. Today we have only very simple video models in our compression systems. Even with the simple models, we can do impressive feats in terms of compression, transformations, object detection, recognition and tracking.

Video is also a heavy consumer of network bandwidth for transmission. One of the main problems in video is how to compress a video stream to fit into a given data pipe. However any act of compression results in a corresponding degradation in video quality. Here lies the basic mystery of compression: We do not exactly know how to measure video quality. It takes us back to the question of how we see and how we perceive the visual world. Though we know that certain kinds of compression are more objectionable than others, we are yet to discover the best compression algorithm optimal in the sense of the largest compression and the least amount of quality loss. We are progressively doing better but a lot remains to be done. Another aspect of the problem is the digital implementation with small size, low power, and low delay. Often we are forced to choose a poor solution because there is no way to implement, make a product and sell it in the mass market. Yet another concern is the backward compatibility with the existing designs. Often international standards organizations come up with an algorithmic solution as the "best" compromise between competing methods.

Almost same situation exists in other areas of video signal processing. In order to do detect objects or recognize faces, we need a good understanding of how we humans do that naturally. This human perceptual model is a good starting point if complimented by other statistical methodologies. There are competing strategies and we need to pick one that should work "today and now".

Today technology is marching faster and faster. It has become easier to design and implement complex algorithms. We are here to help you in your quest for market capture. We have decades of experience in this field and would be very happy to be your guide in this journey.

We have significant experience in video signal processing technologies from the simple to the most complex cutting-edge technologies:

Video Compression Algorithms: Both standards based and the new ones
Pre and Post-processing
Transcoding, scaling and format conversions
Enhancement algorithms
Error concealment algorithms
Multi-camera fusion and metrology
Face/Object Detection and recognition
Deblocking, Denoising, Rate control, and Color Space Management
Geometry processing for sensor and display devices
Compensation for the non-ideal characteristics of display hardware
Compensation for color, brightness, contrast properties of displays
Entropy coder designs for Video