Sonic IR imaging is an emerging NDE technology. This technology uses short pulses of ultrasonic excitation together with infrared imaging to detect defects in materials and structures. Sonic energy is coupled to the specimen under inspection by means of direct contact between the transducer tip and the specimen at some convenient point. This region which is normally in the field of view of the camera appears as intensity peak in the image which might be misinterpreted as defects or obscure the detection and/or extraction of the defect signals in the proximity of the contact region. Moreover, certain defects may have very small heat signature or being buried in noise. In this paper, we present algorithms to improve defect extraction and suppression of undesired heat patterns in sonic IR images. Two approaches are presented, each fits to a specific category of sonic IR images.