jzagaja Posted July 12 Share Posted July 12 I’m developing a video recorder that can capture both uncompressed and compressed RAW, while also recording a flat H.265 stream with minimal ISP processing. The application can process RAW data and encode the result to H.265. It adjusts the saturation vectors in the linear domain to preserve accurate color relationships and supports color targets for calculating a 3×3 color correction matrix and TRCs. It also includes USM and MTF-based filtering in a mildly nonlinear domain, allowing either sharpening or attenuation of high spatial frequencies. I’m also planning to add film simulations after calibrating the camera using manufacturers’ spectral sensitivity curves. I’m open to suggestions regarding additional features worth implementing. At the moment, my main problem is dropped frames during compressed RAW recording. Below is an example from a Motorola Edge 50 Pro: the brighter image is produced by the phone’s ISP, while the other image is processed from RAW. ISP: RAW: RAW: eatstoomuchjam and Cosimo 2 Link to comment Share on other sites More sharing options...
jzagaja Posted July 13 Author Share Posted July 13 Here are a few samples after IT8 color calibration: https://drive.google.com/drive/folders/1EqZM7ZMvsTiZBEeMZ6E6lWjT5Hbzwnrc I’m still working on the in-phone RAW processing algorithms. Cosimo 1 Link to comment Share on other sites More sharing options...
jzagaja Posted July 19 Author Share Posted July 19 I used Ektar drumscan: To create simulation profile: Cosimo 1 Link to comment Share on other sites More sharing options...
jzagaja Posted July 22 Author Share Posted July 22 Correcting a bit IT8 grid. Color checker video V2 target is expensive not sure if implement. Link to comment Share on other sites More sharing options...
jzagaja Posted July 22 Author Share Posted July 22 I like Ektar 100 simulation implemented. Link to comment Share on other sites More sharing options...
techie Posted Friday at 07:59 AM Share Posted Friday at 07:59 AM Compressed RAW is probably the part I'd focus on before adding more film simulations. For the dropped frames, a small diagnostics page could be useful: sensor timestamp delta, queue depth, compression time per frame, sustained write MB/s and dropped-frame count. That should make it much easier to see whether the bottleneck is capture, compression or storage. I'd also preserve per-frame black/white levels and WB/colour-matrix metadata where the device exposes them. Especially useful if you want the RAW processing and calibration to remain consistent across ISO changes. Interesting project. greenscreen and Emanuel 2 Link to comment Share on other sites More sharing options...
jzagaja Posted 1 hour ago Author Share Posted 1 hour ago Right now I’m working on a GPU-accelerated H.265 mode, with the goal of getting video that is as close to RAW as possible while maintaining stable IT8-based color calibration. In RAW mode I’m currently limited to around 22 fps (Motorola Edge 50 Pro). To push beyond that, I’ll probably need to rewrite a significant part of the capture and processing pipeline in C++/NDK and reduce the amount of work done on the Kotlin side. Over the next few days I’m planning to register as a Google Play developer and try to publish an early version of RawLogCam on the Play Store. I’m also considering releasing it first as a closed or open beta. If anyone here has an Android phone with decent Camera2/RAW support and is interested in testing it, I’d definitely appreciate feedback — especially on dropped frames, sustained recording, external SSD recording, RAW compatibility, and color calibration across different sensors. If you have a better idea than going directly through the Play Store for the first public builds, I’m open to suggestions. Link to comment Share on other sites More sharing options...
Recommended Posts
Create an account or sign in to comment
You need to be a member in order to leave a comment
Create an account
Sign up for a new account in our community. It's easy!
Register a new accountSign in
Already have an account? Sign in here.
Sign In Now