Hi all,

I am Mohit, a GSoC student participant with BRL-CAD for the year 2013. My GSoC-13 project was aimed to consolidate and develop an image processing and conversion library. 

To start my work, I adopted BRL-CAD's standard coding style, grasped functions related to its utility library (libbu) and focused on the status of image processing library previously in place (LIBICV). After which I worked on the various tools of Image Processing in BRL-CAD and listed and grouped them as per the processing they are involved in. Prior to my application period and during very initial phase of coding period, I created few patches to earn my commit rights. And was successfully granted one on 26th July.

My first challenge was to decide on the image containers. After discussions on mails and IRC. And writing some testing codes I along with my mentors decided on the current image processing container. This included changing the data type of the image data. Included new entries for number of channels and other image rules. And the most important scrapping the file descriptor used in the image containers. Thus the name also changed from icv_image_file to icv_image. The data was no longer associated to a file but rather an image.

Even bigger challenge was to incorporate these changes in the existing use of the image processing library in the BRL-CAD utilities like rt and libraries like ged and all. Also, I was strictly guided to adopt sequential and minimal change during commits. This required exploring code base of rt and other places where LIBICV was used. Later I completed the task with few modifications and reiterations. This was challenging because rt follows RTUIF (Ray-Tracing User Interface Framework ) involving many files different functions and different privileges to use image reading/writing. With few complications I was later successful in implementing the changes.

After coming to a standard implementation and setting the game rules right other tasks were both easy and smooth. I wrote APIs for image processing of functions including but not limited to convolution (also special convolution including three images [icv_filter3] ), cropping (a special cropping with any four points of a quadrilateral [icv_crop]), resizing (with two ways to interpolate and two to decimate).

I also used these APIs in BRL-CAD's image processing tools and modified there documentation pages when required.

Later I wrote the reading and writting plugin for different image formats. Currently ICV has capibility to read/write images from/to both pipes and files. Also In the case of BRL-CAD raw data of BW, PIX and DPIX the library has an added capability to read without the size being specified. Other implemented format is ppm.

Sean and Erik were great mentors. They would always guide and insist me on adopting standard practices whose importance I realize now. Really maintaining large and old code-bases is both fun and learning.

It was my first time in Open-Source Coding. I enjoyed and believe this summer will be cherishable.

*Future Work*
In future I will be working on other formats. The whole point on adopting the new image container was to use rt's capacities to render high quality images and get an ability to store images of high definition format like OpenEXR in BRL-CAD. Therefore I will be working on it.

--

Mohit

IRC Nick : zero_level