Quantitative scratch assay is significant in cell motility study for tissue repair, evolution of disease, drug treatment, and cancer metastasis. To overcome challenges in traditional manual operations in scratch assay, computational scratch assay is introduced, where image processing algorithms are exploited for cell motility quantification. In this new research realm, dedicated analysis tools are under-developed, which provides many opportunities for researchers expert on signal processing. This work presents a preliminary study in multi-cellular segmentation, which aims to divide a scratch image into wound area and cell-populated regions. The proposed segmentation algorithm consists of a novel LBP-variant edge detector and a parallel processing pipeline. Experimentation on public scratch image benchmark demonstrates the superiority of the proposed method over prior arts. Particularly, the LBP-variant edge detector is capable of generating a single directional-aware edge map so that multiple edge maps along different orientations can be retrieved from it. Taking our preliminary study on multi-cellular segmentation as an example, it is suggested that with carefully designed image processing algorithms, current scratch assay quantification can be much improved.