In most current methods for evaluating the cardiac function by ultrasound, the heart wall area is identified manually by an examiner. To eliminate examiner dependence and to improve usability, an automatic heart wall identification method is desirable. Identification based on only echogenicity often fails because of low echogenicity of some areas of the heart wall. In the present study, to determine more essential features, we focused on the relative temporal change of ultrasonic scatterer distribution and proposed three features for identification of the heart wall and the chamber: cross-correlation of RF signals, that of envelopes, and spatial dispersion of movement vectors in small regions. In an in vivo experiment, using echogenicity and the three features, we identified the heart wall and the chamber in the left ventricular long-axis view, resulting in criteria of separability J of 1.69, 1.40, and 3.02 using these features compared with the result of 0.979 using echogenicity.