Simultaneous comparative evaluation of parameters

Comparative evaluation by applying multiple parameters to a single input video

When developing algorithms, comparative evaluation of different internal parameters is indispensable for measuring the effectiveness of algorithms.

In the past, it was necessary to change the parameters in detail, save the evaluation image for each applied parameter, and generate another comparison image as needed while recording the results. This was due to external constraints on development, such as restrictions on the computational process of providing different parameters and the time required to implement a mechanism to obtain multiple outputs.

A customer used the MIMO DSP Platform to generate four output videos in real time on a GPU, using one video input and providing different parameters. By simultaneously outputting each of these four videos to different video output devices, an environment was created in which parameters can be immediately compared in video images.

By doing so, we were able to reduce the generation process to image files and the process of processing the images to make them comparable, which was previously necessary, and succeeded in dramatically increasing the development efficiency. The ability to evaluate video images as-is has had a greater impact than expected, including the discovery of unexpected characteristics of superior algorithms and the identification of new issues.

Customer Comment: “When we came up with ideas for algorithms that were not available anywhere else in the world, it took a lot of time to apply them to video images. With the MIMO DSP Platform, our development process is much faster, and most of our ideas can be implemented immediately. We feel that this environment has greatly motivated our researchers and accelerated the research process throughout our organization.”

The customer is preparing to use the development time saved by using the MIMO DSP Platform to advance the organization’s new research. Not only are they seeing direct results, but they are also seeing indirect ripple effects in other projects.