Rehabilitation of lower limbs is equally as important as that of upper limbs. This paper presented a study to detect motor imagery of walking (MI-Walking) from background idle state. Broad overlapping neuronal networks involved in reorganization following motor imagery introduce redundancy. We hypothesized that MI-Walking could be robustly detected by constraining dependency among selected features and class separations. Hence, we proposed to jointly select channels and frequency bands involved in MI-Walking by optimizing/regularizing the objective function formulated on the dependency between features and class labels, redundancy between to-be-selected with selected features, and separations between classes, namely, “regularized maximum dependency with minimum redundancy-based joint channel and frequency band selection (RMDR-JCFS)”. Evaluated on electroencephalography (EEG) data of 11 healthy subjects, the results showed that the selected channels were mainly located at premotor cortex, mid-central area overlaying supplementary motor area (SMA), prefrontal cortex, foot area sensory cortex and leg and arm sensorimotor representation area. Broad frequencies of alpha, mu and beta rhythms were involved. Our proposed method yielded an averaged accuracy of 76.67%, which was 9.08%, 5.03%, 7.03%, 14.15% and 3.88% higher than that obtained by common spatial pattern (CSP), filter-bank CSP, sliding window discriminate CSP, filter-bank power and maximum dependency and minimum redundancy methods, respectively. Further, our method yielded significantly superior performance compared with other channel selection methods, and it yielded an averaged session-to-session accuracy of 70.14%. These results demonstrated the potentials of detecting MI-Walking using proposed method for stroke rehabilitation.