Ultrasonic backscatter has been gradually applied to the assessment and diagnosis of bone disease. The heavy frequency-dependent attenuation of ultrasound results in weak ultrasonic signals with poor signal-to-noise ratio and serious wave distortions during propagation in cancellous bone. Ultrasonic attenuation measured with the through-transmission method is an averaged result of ultrasonically interrogated tissues (including the soft tissue, cortical bone and cancellous bone). Therefore, the through-transmission measurements can not accurately provide ultrasonic attenuation of cancellous bone of interest. The purpose of this study is to estimate ultrasonic frequency-dependent attenuation with ultrasonic backscatter measurements and to compensate for the frequency-dependent attenuation in an ultrasonic backscatter signal from cancellous bone. In-vitroultrasonic backscatter and through-transmission measurements are performed on 16 cancellous bone specimens by using 1.0-MHz transducers. Spatial scans are performed in a 10 mm × 10 mm scanned region with a spatial interval of 0.5 mm for each bone specimen. The frequency slope of ultrasonic attenuation is measured with the ultrasonic through-transmission signals serving as a standard value. Four different algorithms (the spectral shift method, the spectral difference method, the spectral log difference method, and the hybrid method) are used to estimate the frequency slope of ultrasonic attenuation coefficient from ultrasonic backscatter signal. The results show that the frequency-dependent attenuation coefficient ranges from 2.3 dB/mm/MHz to 6.2 dB/mm/MHz for the bovine bone specimens. The through-transmission measured frequency slope of ultrasonic attenuation coefficient is (4.14 ± 1.14) dB/mm/MHz (mean ± standard deviation), and frequency slopes of ultrasonic attenuation coefficient are estimated by four backscattering methods to be (3.88 ± 1.15) dB/mm/MHz, (4.00 ± 0.98) dB/mm/MHz, (3.77 ± 0.84) dB/mm/MHz, and (4.05 ± 0.85) dB/mm/MHz, respectively. The estimated frequency-dependent attenuation is significantly correlated with the standard attenuation value ( R= 0.78－0.92, p< 0.01), in which the spectral difference method ( R= 0.91, p< 0.01) and the hybrid method ( R= 0.92, p< 0.01) are more accurate with an estimated error less than 20%. The results prove that it is feasible to measure the frequency-dependent attenuation from ultrasonic backscatter signal of cancellous bone. Based on Fourier transform-inverse Fourier transform, the frequency-dependent attenuation can be compensated.The compensated ultrasonic signals are with significantly improved signal intensity and improved signal-to-noise ratio. This study is conducive to the subsequent ultrasonic backscatter measurement and ultrasonic imaging of cancellous bone.