The goal of this study was to develop, validate, and implement an image analysis framework to automatically analyze chondrocytes in 3D image stacks of cartilage acquired using a fluorescent confocal microscope. Source specimens consist of viable osteochondral tissue co-stained with multiple live-cell dyes. Our framework utilizes a seeded watershed-based algorithm to automatically segment individual chondrocytes in each 2D slice of the confocal image stack. The resulting cell segmentations are colocalized in 3D to eliminate duplicate segmentation of the same cell resulting from the visibility of fluorescence signal in multiple imaging planes, and the 3D cell distribution is used to automatically define the cartilage tissue volume. The algorithm then provides chondrocyte density data, and the associated segmentation can be used as a mask to extract and quantify per cell intensity of a secondary, functional dye co-staining the chondrocytes. The accuracy of the automated chondrocyte segmentation was validated against manual segmentations (average IOU = 0.79). When applied to a cartilage surrogate, this analysis framework estimated chondrocyte density within 10% of the true density and demonstrated a good agreement between framework's counts and manual counts (R² = 0.99). In a real application, the framework was able to detect the increased dye signal of monochlorobimane (MCB) in chondrocytes treated with N-acetylcysteine (NAC) after mechanical injury, quantifying intracellular biochemical changes in living cells. This new framework allows for fast and accurate quantification of intracellular activities of chondrocytes, and it can be adapted for broader application in many imaging and treatment modalities, including therapeutic OA research.