The possibility of applying adaptive correction to ground-based solar astronomy is considered. Different ways of development of an adaptive correction to be used in the BSVT of the Baikal Astrophysical Observatory are discussed. The correlation technique for measuring the displacement of an image fragment consists in the following: a reference frame is stored and then the correlation function of irradiance distribution in the reference frame and any pre-selected current frame is calculated. The position of the maximum of the correlation function determines the coordinates of displacement of the current frame with respect to the reference one. The correlation tracker failed to track the image, because, as turned out, the CCD array of the DALSTAR camera featured four equidistant horizontal linear arrays with the sensitivity 0.4-1.2% lower than that of the other elements, which led to the appearance of four lines in the recorded image. With the uniform irradiance of the array, the contrast of these lines was 0.2-0.6%. When low-contrast objects were observed, the static local maximum could appear higher than the moving one associated with the image displacement. In this case, the tracker failed to detect the image motion. During the tests of the adaptive optical system at the BSVT in August 2003 and August 2004, the contrast of the granulation pattern in different areas of the solar disk (at the center, at an edge, near sunspots) averaged from 1 to 4%. To calculate the mutual correlation function of the irradiance distribution in the reference and current frames, out algorithm employs the mixed-radix Fast Fourier Transform (FFT). The algorithm operates in the 16-bit floating point arithmetic. When recording displacements of the granulation pattern, we noticed that the mutual correlation function of the reference and current frames had several extremes, in particular, static ones, attributed to the detector features. To separate the maximum of the correlation function, whose coordinates determine the displacement of the granulation pattern, the following algorithm is proposed. In the spatial spectrum of the image of the reference and current frames, the characteristic frequencies associated with the granule size are amplified. The amplification contour is a gaussian, whose parameters are defined for every detected fragment of the granulation pattern. Thus, with the properly chosen parameters, low frequencies are suppressed in the spatial spectrum, which leads to smoothing of the irradiance over the detector field of view, and high frequencies associated, in particular, with defects of the detector array are suppressed as well. Once these spectra are multiplied and the inverse Fourier transform is calculated, the maximum of the correlation function, attributed to the shift of the solar granulation pattern intensifies. The use of the movies allowed us to find certain criteria for selection of the parameters of the optimal amplification contour. The correctness of measurement of the image displacement by the modified correlation tracker (MCT) was checked by comparison with the displacement measured by the traditional correlation tracker (TCT). For this check, the image fragments, which are successfully processed by the traditional correlation tracker, were taken.The contrast of the spot was 21%. The rms deviations of the tracking signals measured by TCT are 0.723 arc sec (along the axis X) and 5.121 arc sec (along the axis Y), for MCT they are 0.720 arc sec and 5.119 arc sec, respectively. The movie was recoded on August 3 of 2004 at the strong wind. In the image jitter spectrum, a maximum at the frequency of 7.6 Hz, associated with the oscillations of the siderostat mirror about the fixation axis, was observed. During the realization, the contrast of the granulation pattern varied from 1.80 to 2.49% with the average value of 2.12%. For the traditional correlation tracker, the maxima of the mutual correlation function, associated with the static structure of the image and the moving one, have close values. The tracker either indicates ``0'' or tries to track the image motion. The mutual correlation function of MCT has a pronounced maximum, which shifts together with the image.The rms deviations of the tracking signals measured by MCT amount to 1.345 arc sec (along the axis X) and 4.495 arc sec (along the axis Y). The values obtained are in a good agreement with the corresponding values for the previous realization. The time interval between the realizations compared was 9 minutes. Hence of the modified correlation tracker algorithm permits the detection of displacements of low-contrast image fragments.