We support two different mode to view the original image and processed image: slider view and parallel side-by-side view. Slider view is implemented by adding two sliders on the right and bottomof the displayed image: one vertical and one horizontal. User can slide them to form four differentregions: top-left, top-right, bottom-left, bottom-right. User can choose one of the four regions to displaythe original image. Notice that we did not put relevant state/variables in the model because the sliderview is local to the image displaying UI and does not involve underlying data such as the image andthe transformations. On the other hand, the parallel view is implemented by just putting two JavaFxImageViewpanel side by side.
Multi-threaded processing: From our experiments, we have noticed performance drops on large images.We have decided to implement multi-threaded processing for some of the operations to reduce the effectof this performance drop. For simple operations such as black & white, flipping images, etc. We simply employed divide and conquering technique whereby we spawn a number of threads, giving each threada portion of the image to work on. More advanced operations require combining intermediate results of each threads before merging into final result.
We support batch processing under the "Batch" tab, which contains the same set of image processing operation as in the single image processing tab. For batch processing, we use an executor pool and pickeach image and operation and work on them in a pipeline fashion
Import/Export sequences of operations: As our users may often need to apply the same sequence oftransformations to a number of different images, we added the ability to export a list of transformationsas a JSON file, so that the same pipeline can be easily applied repeatably at a later time. As all of our transformations already all implement the ImageProcessing interface, this involved declaring them as Serializable, giving them an unique SerialName, and finally marking which fields needs to be includedin the JSON file (if necessary). Exporting the operations is then just using Kotlin’s JsonBuilder to write the list of operations as a JSON file. Importing of operations involves decoding the JSON file as alist of objects, then applying the operations one after another.
See the project report for more details.