This paper presents a new technique for electroencephalography (EEG) seizure detection from multi-channel EEG signals based on image processing concepts in Internet-of-Health-Things (IoHT) systems. The multi-channel EEG segments are treated as two-dimensional matrices as if they were images. Scale-space analysis with Scale Invariant Feature Transform (SIFT) is used to extract the feature points in the up-mentioned two-dimensional matrices. The number of points is used as a discriminating factor between seizure segments and normal segments. An exhaustive study of the 24 patients of the Children's Hospital Boston (CHB-MIT) database is presented in this paper. The EEG signals are transmitted via WiFi/Bluetooth, then all their signals are segmented into one-second segments, the numbers of features points are extracted from these segments, the Probability Density Function (PDF) of the number of feature points for normal and seizure segments are estimated. The Equal Error Rate (EER) is estimated between PDFs of the numbers of feature points in seizure and normal segments. Simulation results on all patients reveal the ability of the proposed technique to set a patient-specific discrimination threshold of 70% of Max spectral power for seizure detection with an accuracy of 95.6%.