Counting the number of rice seeds is essential in assessing the quality of rice varieties such as yield, rice diseases (sloppy rice), etc. Besides the machine learning approach, there are disadvantages. Contrary to machine learning, the
counting approach utilizes image processing. This research contributes to counting the number of rice grains per panicle by combining contrast limited adaptive histogram equalization and Candy algorithm on a data set of 150 rice samples. The method is conducted through the steps of denoising, converting RGB color channels to LAB, image segmentation and contouring, and finally counting. The study’s results were evaluated by comparing counting by hand and by algorithm; the error result was ~0.902%.