To detect the particle agglomeration degree for assessing crystal growth quality during a crystallization process, an in situ image analysis method is proposed based on a microscopic double-view imaging system. First, a fast image preprocessing approach is adopted for segmenting raw images taken simultaneously from two cameras installed at different angles, to reduce the influence from uneven illumination background and solution turbulence. By defining an index of the inner distance based curvature for different particle shapes, a preliminary sieving algorithm is then used to identify candidate agglomerates. By introducing two texture descriptors for pattern recognition, a feature matching algorithm is subsequently developed to recognize pseudoagglomerates in each pair of the double-view images. Finally, a fast algorithm is proposed to,count the number of recognized particles in these agglomerates, besides the unagglomerated particles. Experimental results from the potassium dihydrogen phosphate (KDP) crystallization process demonstrate good accuracy for recognizing pseudoagglomeration and counting the primary particles in these agglomerates by using the proposed method.