Particle size and shape significantly affect powder processing and their end-product quality in a variety of industries. Imaging methods can successfully characterize populations of needle-like particles. Prior to off-line imaging, adjusting the particle density can reduce particle overlaps but increase measurement/processing times. Discarding data of overlapping particles, as most image processing algorithms do, biases particle size and shape distributions. Building on previous efforts, we here provide an image processing technique that accurately separates and sizes overlapping needle-like particles. Our algorithm combines edge detection, layer-stripping watershed segmentation and length approximation. To test the algorithm, a large number of real particle projections were randomly overlapped with various overlap intensities. Approximately 92–72% of the particles were detected and the particles’ dimensions were characterized with an accuracy of 87–75%, with these ranges corresponding to low and high overlap intensities. Overall, the algorithm removes biases to considerably improve characterization accuracy of powders containing needle-like particles.